Local membrane charge regulates β₂ adrenergic receptor coupling to G_i3

Abstract

The β₂ adrenergic receptor (β₂AR) signals through both G_s and G_i in cardiac myocytes, and the G_i pathway counteracts the G_s pathway. However, G_i coupling is much less efficient than G_s coupling in most cell-based and biochemical assays, making it difficult to study β₂AR−G_i interactions. Here we investigate the role of phospholipid composition on G_s and G_i coupling. While negatively charged phospholipids are known to enhance agonist affinity and stabilize an active state of the β₂AR, we find that they impair coupling to G_i3 and facilitate coupling to G_s. Positively charged Ca²⁺ and Mg²⁺, known to interact with the negative charge on phospholipids, facilitates G_i3 coupling. Mutational analysis suggests that Ca²⁺ coordinates an interaction between phospholipid and the negatively charged EDGE motif on the amino terminal helix of G_i3. Taken together, our observations suggest that local membrane charge modulates the interaction between β₂AR and competing G protein subtypes.

In the healthy heart, the β₂ adrenergic receptor (β₂AR) signals through G_s and G_i proteins but the mechanism underlying G protein selectivity is not fully understood. Here, the authors show that membrane charge and intracellular cations modulate the β₂AR−G_i3 interaction.

Introduction

A third of all FDA-approved pharmaceutical drugs function by modulating the activity of G-protein-coupled receptors (GPCRs)¹, a large receptor superfamily. GPCRs catalyze the activation of heterotrimeric G proteins, which in turn initiate a multitude of signaling cascades that alter cellular function.

The β₂ adrenergic receptor (β₂AR) is a prototypical GPCR that mediates the fight-or-flight response. β₂AR signals through both G_s and G_i, and the dual G protein selectivity of β₂AR is best characterized in heart muscle (cardiac myocytes). In healthy neonatal cardiac myocytes, epinephrine-stimulated β₂AR immediately activates G_s, increasing contraction rate, but after 10−15 min β₂AR signals predominantly through G_i ², which decreases the contraction rate. In cardiac myocytes, β₂AR couples to both G_i2 and G_i3³. Of interest, G_i activation is impaired if β₂AR internalization is blocked⁴. Also, G_i does not interact with a modified β₂AR that internalizes but does not recycle to the plasma membrane⁵, or with WT β₂AR that internalizes but is pharmacologically blocked from recycling⁶. Taken together, these observations demonstrate that β₂AR−G_i interaction is regulated temporally and perhaps spatially.

During heart failure, a condition of chronic, progressive cardiac insufficiency, the G_i pathway counteracts some negative consequences of chronic G_s activation that exacerbate heart failure, namely apoptosis and structural and functional remodeling^7–9. However, G_i activation reduces contractility, which can be problematic in certain models of heart failure¹⁰. More precise regulation of G_s and G_i activation is a therapeutic aim⁹.

While β₂AR signals through both G_s and G_i, the mechanism that initiates the G_s-to-G_i switch in the healthy heart is not fully understood. Multiple biochemical mechanisms may play a role. PKA phosphorylation of β₂AR has been reported to increase G_i1 coupling in vitro¹¹ and G_i coupling in HEK cells¹²; however, the G_i1 subtype is not expressed in cardiac myocytes, and β₂AR−G_i coupling is PKA independent in these cells². In addition, GRK2 phosphorylation of β₂AR has been suggested to increase G_i coupling¹³, but other investigators have reported that dephosphorylation is critical for β₂AR recycling to the plasma membrane, and β₂AR−G_i interactions⁶. Therefore, the mechanisms that modulate β₂AR−G_i coupling remain unclear.

In vitro, negatively charged phospholipids stabilize an active conformation of the β₂AR and enhance its affinity for the catecholamine isoproterenol¹⁴, but the effect of phospholipid charge on G protein coupling is unknown. Negatively charged lipids have previously been shown to facilitate β₁AR−G_s interaction¹⁵, NTS1−G_q¹⁶ interaction, CB₂−G protein interaction¹⁷, and rhodopsin−G_t¹⁸ interaction. In cardiac myocytes, β₂AR activates G_s in T-tubules¹⁹, deep evaginations of the plasma membrane enriched in L-type calcium channels and negatively charged phospholipids²⁰. β₂AR activation of G_i may also occur in T-tubules, after internalization and recycling. However, trafficking events may alter the composition of phospholipids surrounding the β₂AR. In addition, the β₂AR−G_s signaling that occurs prior to β₂AR−G_i signaling greatly increases the Ca²⁺ concentration near T-tubules²¹, which may alter the charge properties of the T-tubule lipids²².

Here we examine the effect of phospholipid charge on β₂AR interactions with G_s and G_i3. We find that negatively charged lipids enhance β₂AR interaction with G_s and impair interaction with G_i3. Further, Ca²⁺ and Mg²⁺ facilitate β₂AR−G_i3 interaction in negatively charged lipids. Our observations suggest that local membrane charge, tuned by intracellular cations, modulates β₂AR interaction with G_i3.

Results

Monitoring G protein coupling by fluorescence spectroscopy

Epinephrine activates β₂AR by partially stabilizing the conformation recognized by the G protein. This conformation is fully stabilized upon G protein coupling^23–25. A feature of the G-protein-coupled conformation of β₂AR is a 14 angstrom outward displacement of the cytoplasmic end of β₂AR transmembrane segment 6 (TM6) (Fig. 1a). This conformation can be detected by fluorescence spectroscopy, using a modified β₂AR labeled at the cytoplasmic end of TM6 with an environmentally sensitive fluorophore, monobromobimane (mB−β₂AR, see Methods and ref. ²⁶). Outward movement of TM6 affects the environmental polarity of mB, going from the hydrophobic receptor core to the solvent, decreasing the intensity of mB and increasing the wavelength where emission intensity is greatest (λ_max), from ~447 to ~468 nm (Fig. 1a). We monitor λ_max to detect G protein coupling, as λ_max increases with coupling (Supplementary Fig. 1).

Fig. 1.

Negatively charged lipids inhibit β₂AR−G_i coupling in detergent. a mB−β₂AR emission spectra in DDM + CHS micelles (5:1 DDM:CHS mole ratio) in the absence of epinephrine (APO) and in the presence of epinephrine, ± G protein (G_s or G_i3). Arrows point to the lambda max value, i.e. the wavelength where mB emission intensity is greatest. Inset shows how G_s coupling alters the structure of mB−β₂AR, highlighting the change in position of monobromobimane (mB) at C265 of transmembrane 6 (TM6) (Inactive: PDB 5JQH⁷⁴; Active: PDB 3SN6⁶⁹). b Structures of CHS and phospholipids, with net charge indicated. c Effect of phospholipid on mB−β₂AR interaction with G protein (G_i1, G_i2, G_i3, and G_s), read out as an increase in lambda max. Interaction was assessed in the absence of lipid (5:1 DDM:CHS mole ratio) and in the presence of POPG, POPS, POPC, or POPE (5:1:1 DDM:CHS:Lipid mole ratio). Multiplicity adjusted P values were computed by two-way ANOVA followed by Dunnett’s post hoc test between indicated groups. d Selected mB−β₂AR emission spectra from panel (c). e Effect of CHS on mB−β₂AR interaction with 1 μM G_i3. Interaction was assessed in the presence of CHS (5:1:1 DDM:POPE:CHS mole ratio) or in the absence of CHS (5:1 DDM:POPE mole ratio). Multiplicity adjusted P values were computed by two-way ANOVA followed by Sidak’s post hoc test between indicated groups. a–e mB−β₂AR concentration is 300 nM. Data are mean of three independent experiments. f β₂AR-induced GTP turnover for G_i3 (± 200 μM epinephrine, EPI) in DDM + POPS (5:1 DDM:POPS mole ratio) and in DDM + POPE (5:1 DDM:POPE mole ratio). Free GTP was assayed after 12 min. Luminescence signals were normalized relative to the condition with G_i3 alone (see Supplementary Fig. 4). Data are mean ± s.e.m. of four independent experiments. Multiplicity adjusted P values were computed by two-way ANOVA followed by Sidak’s post hoc test between indicated groups. Source data are provided in the Source Data File

Negatively charged lipids inhibit β₂AR-G_i coupling

In the presence of epinephrine, we observe a change in intensity and λ_max of mB−β₂AR following the addition of G_s in a detergent mixture containing n-dodecyl-β-d-maltopyranoside (DDM) and cholesteryl hemisuccinate (CHS) that is commonly used for biochemical study of GPCR/G protein complexes (Fig. 1a). In contrast, the coupling efficiency of mB−β₂AR−G_i3 was relatively weak (Fig. 1a). Next, we compared the coupling efficiency of mB−β₂AR−G_i in DDM + CHS mixtures with different phospholipids incorporated (Fig. 1b–d). While we were unable to detect interactions of mB−β₂AR with G_i1, G_i2 or G_i3 in the presence of negatively charged lipids phosphatidylserine (POPS) and phosphatidylglycerol (POPG), we observed a weak interaction with G_i1 and G_i3 in neutral lipids phosphatidylethanolamine (POPE) and phosphatidylcholine (POPC) (Fig. 1c, d—G_i1 and G_i3 with epinephrine stabilizes more mB−β₂AR in the active conformation than epinephrine alone). This result suggested that negatively charged lipids may repel G_i1 and G_i3 interaction with β₂AR.

Indeed, we observed that negatively charged CHS decreased mB-β₂AR-G_i3 coupling (Fig. 1e), supporting our hypothesis that negatively charged lipids inhibit coupling. Since CHS is a nonphysiologic cholesterol analog, we omitted negatively charged CHS in subsequent experiments in order to isolate the effect of phospholipid charge on mB−β₂AR−G_i interactions.

Given reports that PKA phosphorylation of β₂AR increases β₂AR−G_i interaction in vitro¹¹, we also tested the effect of PKA phosphorylation, but no enhancement of mB−β₂AR interaction with G_i1, G_i2, or G_i3 was observed (Supplementary Fig. 2), suggesting that phosphorylation does not potentiate β₂AR−G_i interaction under our experimental conditions, and that other mechanisms may enhance β₂AR−G_i interaction.

In contrast to G_i1 and G_i3, we were unable to detect coupling to G_i2 in any of the lipid-detergent mixtures (Fig. 1c), suggesting that the interaction is low affinity. Of interest, G_i1 and G_i3 share higher sequence identity than either subtype shares with G_i2. To confirm the lack of coupling to G_i2 observed by fluorescence analysis, we examined β₂AR-stimulated G protein turnover of GTP in net neutral DDM + POPC for G_i2 and G_i3 (Supplementary Fig. 3). As expected from our fluorescence studies, we observed robust β₂AR-induced GTP turnover for G_i3, but only weak β₂AR-induced GTP turnover for G_i2. In contrast we observed strong coupling of G_i2 to purified neurotensin receptor, indicating that the G_i2 protein was functional. While both G_i2 and G_i3 couple to β₂AR in cardiac myocytes³, the coupling efficiency using purified components is substantially different (see Discussion). We therefore focused on β₂AR−G_i3 coupling in further studies.

Next we examined how the lipid environment affects β₂AR activation of G_i3. In the presence of epinephrine, significantly more β₂AR-induced GTP turnover was detected in DDM micelles containing POPE (net neutral lipid) than in DDM micelles containing POPS (net negative) (Fig. 1f). This effect on β₂AR-mediated turnover was significant, even though the lipid:DDM mole ratio was only 1:5. POPE did not increase basal GTP turnover by G_i3 in the absence of β₂AR (Supplementary Fig. 4). Taken together, these results indicate that the charge property of phospholipids regulates G_i activation by β₂AR.

Ca²⁺ facilitates G_i3 coupling in negatively charged lipids

Ca²⁺, a ubiquitous second messenger, plays an important role in cardiac myocytes; Ca²⁺ waves, magnified by G_s activation, drive the cardiac myocyte contraction machinery. Recently Ca²⁺ was reported to regulate T-cell receptor activation by modulating the charge property of lipids²⁷. Given that Ca²⁺ interaction with negative charge on phospholipids neutralizes the charge, we tested whether Ca²⁺ improves mB−β₂AR−G_i3 coupling efficiency in negatively charged DDM + POPS. Indeed, Ca²⁺ improved coupling efficiency in DDM + POPS micelles (Fig. 2a), and this effect required lipid (Fig. 2b). Moreover, Ca²⁺ had little effect on mB−β₂AR−G_s interaction, implicating differences in G_s and G_i3 surface charge.

Fig. 2.

Ca²⁺ facilitates G_i3 coupling in negatively charged phosphatidylserine. a The effect of 10 mM CaCl₂ on mB−β₂AR−G protein interaction (G_i3 and G_s) was examined in micelles containing 5:1 DDM:POPS (mole ratio). Data were collected in the absence (APO) and presence (100 μM) of epinephrine. mB−β₂AR concentration is 250 nM. Data are mean ± s.e.m. of three independent experiments. Multiplicity adjusted P values were computed by two-way ANOVA followed by Sidak’s post hoc test between CaCl₂ conditions. b The effect of 10 mM CaCl₂ on mB−β₂AR−G_i3 interaction in DDM micelles ± POPS (i.e. DDM alone vs. 2.5:1 DDM:POPS mole ratio). mB−β₂AR concentration is 250 nM. Data are mean ± s.e.m. of three independent experiments. Multiplicity adjusted P values were computed by three-way ANOVA followed by Sidak’s post hoc test between indicated groups. Source data are provided in the Source Data File

Ca²⁺ interacts with the amino terminal helix of G_i3

Next, we sought to determine the mechanism by which Ca²⁺−POPS interactions increase mB−β₂AR coupling to G_i3 but not to G_s. Given that the amino terminal helix (αN) of G protein is adjacent to the membrane when coupled to the β₂AR²⁵, and polybasic residues on G_s αN are known to facilitate membrane interaction²⁸, we looked for a possible selectivity determinant within αN. Since αN of G_s and G_i are differentially charged (Fig. 3a), we first replaced αN of G_s with αN of G_i3, creating a G_i3-G_s chimera (Fig. 3b).

Fig. 3.

Ca²⁺ interacts with the amino terminal helix of G_i3. a Membrane-facing surfaces of G_s and G_i1 Ras domains, with β₂AR shown in gray. The membrane-facing surface of G_i1 was modeled by superimposing the Ras domain of G_i1 (PDB: 1GP2⁷⁰) onto the structure of G_s in complex with β₂AR (PDB: 3SN6⁶⁹). Red and blue signify negative and positive charge on G protein, respectively. Dashed lines highlight the region in αN where charge differs: The sequence is KDKQ in WT G_s vs. EDGE in WT G_i1 (and in G_i2, G_i3). b mB−β₂AR−G protein dose−response curves ± 10 mM CaCl₂. Data were generated with the G protein depicted below the curves: mutations were made in αN and corresponding electrostatic models are shown. Epinephrine was not included in experiments titrating G_s WT, G_i3-G_s Chimera, or G_s-neg. to enhance the effect of CaCl₂. Epinephrine (100 μM) was included in experiments titrating G_i3 WT and G_i3-pos. mB−β₂AR concentration is 250 nM. Data are mean ± s.e.m. of three independent experiments. Multiplicity adjusted P values were computed by two-way ANOVA followed by Sidak’s post hoc test between CaCl₂ conditions. Source data are provided in the Source Data File

While Ca²⁺ does not promote mB−β₂AR coupling to WT G_s (Fig. 3b), it did promote mB−β₂AR coupling to the G_i3-G_s chimera (Fig. 3b). Next, we compared the membrane-facing charge of G_s WT αN and G_i3 WT αN. Structural analysis revealed that charge differed at the C terminal end of αN: G_i harbors a negatively charged motif (EDGE) at the position where G_s harbors a positively charged motif (KDKQ) (Fig. 3a). To examine whether this motif dictates a differential response to Ca²⁺, we constructed a G_s mutant (G_s-neg.) containing the negatively charged motif of G_i3 (KDKQ → EDGE). Ca²⁺ increased mB−β₂AR interaction with this mutant (Fig. 3b), suggesting the EDGE motif is responsible for the effect of Ca²⁺ on G_i3 αN. Taken together, our results imply that Ca²⁺ coordinates an interaction between the negatively charged EDGE motif on αN of G_i3 and the headgroup of POPS. In the absence of Ca²⁺, like-charge repulsion decreases mB-β₂AR coupling to G_i3.

We also constructed a G_i3 mutant (G_i3-pos.) containing the positively charged motif of G_s (EDGE → KDKQ). The mutations only partially removed the effect of Ca²⁺ (Fig. 3b), indicating the effect of Ca²⁺ on G_i3 extends beyond an effect on αN (see Discussion).

Bilayer charge differentially affects G_s and G_i3 coupling

Owing to their geometry and charge, phospholipids might induce changes in the size and shape of micelle assemblies which could also influence mB−β₂AR−G protein interaction^29,30. To examine the effects of phospholipids in a more native environment, and to restrict the size and shape of lipid ensembles, helping us isolate the effect of membrane charge, we reconstituted mB−β₂AR into nanodisc bilayers and purified the nanodiscs to homogeneity using size-exclusion chromatography (Supplementary Fig. 5).

First, we compared the influence of lipid composition in the absence of Ca²⁺. Negatively charged bilayers (DOPG and DOPS bilayers), previously reported to increase agonist affinity¹⁴, expectedly red-shifted the emission spectra of unliganded mB−β₂AR, suggesting these lipids stabilize an active conformation. Additionally, unliganded mB-β₂AR and G_s could fully couple in negatively charged bilayers but not in neutral bilayers (Fig. 4), suggesting negatively charged lipids might facilitate signaling through G_s. In contrast, negatively charged lipids (especially DOPS bilayers) decreased G_i3 coupling to epinephrine-activated mB−β₂AR (Fig. 4).

Fig. 4.

Bilayer charge differentially affects G_s and G_i3 coupling. The effect of G_s (left) and G_i3 (right) concentration on mB−β₂AR fluorescence was examined in nanodisc bilayers of varying phospholipid composition (DOPC, DOPE, DOPG, DOPS). Epinephrine was omitted (APO) or included (30 μM). Interaction with G protein is read out as an increase in lambda max. The net charge of the phospholipid is indicated in parentheses. mB−β₂AR concentration is 100 nM; maximum stoichiometry is 12:1 (for G_s) and 24:1 (for G_i3). Data are mean ± s.e.m. of three independent experiments. Source data are provided in the Source Data File

In fact, in negatively charged bilayers without epinephrine, G_i3 unexpectedly blue-shifted the emission spectra of mB−β₂AR. While this may indicate that G_i3 stabilizes the β₂AR in an inactive conformation in negatively charged lipids, it may represent a nonspecific interaction of inactive G_i3 with the β₂AR or the lipid bilayer.

Cations promote G_i3 coupling in negatively charged bilayers

Next we examined the effect of Ca²⁺ and Mg²⁺. In the absence of G_i3, both Ca²⁺ and Mg²⁺ reversed the active-state stabilizing effect of negatively charged DOPS and DOPG bilayers (Fig. 5a). Despite this, Ca²⁺ and Mg²⁺ increased mB-β₂AR coupling to G_i3 in negatively charged DOPS bilayers, but only Ca²⁺ was efficacious at concentrations below 1 mM (Fig. 5a). Ca²⁺ similarly affected mB−β₂AR−G_i3 interaction in negatively charged DOPG bilayers (Fig. 5a), but the magnitude of the effect in DOPG bilayers was less than observed in DOPS bilayers due to the higher baseline effect of DOPG on β₂AR conformation.

Fig. 5.

Ca²⁺ and Mg²⁺ facilitate G_i3 coupling in negatively charged bilayers a The effect of CaCl₂ and MgCl₂ concentration on mB−β₂AR fluorescence in DOPS and DOPG nanodisc bilayers was examined in the presence and absence of G_i3. Epinephrine was included (30 μM) or omitted (APO). EC₅₀ is mean ± s.e.m. b The effect of G protein concentration (G_i3 and G_s) on mB−β₂AR fluorescence ± 3 mM CaCl₂ was examined in DOPS nanodiscs in the absence (APO) and in the presence (EPI) of epinephrine (30 μM). c The effect of G_i3 on mB−β₂AR fluorescence ± 3 mM CaCl₂ was examined in DOPE, DOPC, and DOPG nanodisc bilayers in the presence of 30 μM epinephrine (EPI). d The effect of G_i2 concentration on mB−β₂AR fluorescence ± 3 mM CaCl₂ was examined in DOPS nanodisc bilayers in the presence of 30 μM epinephrine (EPI). a–d mB−β₂AR concentration is 100 nM. The net charge of the phospholipid molecule is indicated in parentheses. Data are mean ± s.e.m. of three independent experiments. Source data are provided in the Source Data File

We also compared the effect of Ca²⁺ on mB−β₂AR interactions with G protein subtypes (G_s versus G_i3) in DOPS bilayers. As observed in micelles, Ca²⁺ increased mB−β₂AR coupling to G_i3 but not to G_s (Fig. 5b). Ca²⁺ also improved mB−β₂AR−G_i3 coupling efficiency in negatively charged DOPG bilayers (Fig. 5c). However, incorporating Ca²⁺ did not enable detection of mB−β₂AR interaction with the G_i2 subtype of G_i (Fig. 5d, see Discussion).

While Ca²⁺ interacts with net negative PS and PG with relatively high affinity, it also interacts with the negatively charged phosphate group on net neutral PC³¹ and PE lipids^31,32. In neutral DOPE and DOPC bilayers (Fig. 5c and Supplementary Fig. 6), Ca²⁺ only slightly enhanced mB−β₂AR−G_i3 interaction, which could be attributable to weaker Ca²⁺/DOPE and Ca²⁺/DOPC interactions that have been reported. Taken together, our observations provide biochemical evidence that local membrane charge can regulate β₂AR−G protein interaction.

Discussion

We observed that local membrane charge regulates β₂AR−G protein interaction. Negatively charged membrane promotes β₂AR−G_s coupling and suppresses β₂AR−G_i3 coupling. However, G_s bias is reduced in neutral membrane and in negatively charged membrane in the presence of divalent cations (see model in Fig. 6).

Fig. 6.

Membrane charge is a tunable modulator of β₂AR−G protein interaction. Models depict epinephrine-bound β₂AR. Left: In negatively charged lipids, β₂AR−G_s coupling is efficient, but β₂AR−G_i3 coupling is relatively inefficient, in part because β₂AR−G_i3 attraction is countered by membrane-G_i3 repulsion. Specifically, negatively charged lipids repel the negatively charged EDGE motif found on the amino terminal helix of G_i3 (shown in red), a region that is positively charged in G_s (shown in blue). Right: Two mechanisms that neutralize membrane charge facilitate β₂AR coupling to G_i3. These mechanisms may play a role in G_s-to-G_i switching in cardiac myocytes. Top right: Ca²⁺ and Mg²⁺ stabilize a like-charge interaction between the membrane and the EDGE motif. (Note that the effect of Ca²⁺ and Mg²⁺ may extend beyond an effect on αN positioning.) Bottom right: Epinephrine-stimulated β₂AR traffics to membrane without negatively charged lipids

We have begun to explore the mechanism by which Ca²⁺ increases mB−β₂AR coupling to G_i3 in phospholipids. The effect of Ca²⁺ was largely dependent on charged groups on G_i3, indicating Ca²⁺ doesn’t simply affect membrane structure. Although G proteins are membrane tethered via lipidation, the lipid anchor of G_i3 is not sufficient for optimal interaction with β₂AR in negatively charged bilayers, possibly due to repulsion of the carboxyl terminal end of the αN helix. We propose that Ca²⁺ helps orient the carboxyl terminal end of the αN helix of G_i3 near the membrane, thereby facilitating β₂AR−G_i3 interactions. More specifically, we propose that Ca²⁺ facilitates the interaction of lipid with the negatively charged EDGE motif on the αN helix of G_i3. Ca²⁺ may coordinate a like-charge interaction between the carboxylate groups on the G_i3 EDGE motif and the phosphate group present on all lipids. In addition, Ca²⁺ may especially stabilize EDGE interaction with PS lipids by coordinating a like-charge interaction between the carboxylate groups on the G_i3 EDGE motif and the carboxylate group on PS (not present on PG) (refer to structures in Fig. 1b), or Ca²⁺ might coordinate an intramolecular interaction between the phosphate group and the carboxylate group on PS, freeing the amino group (NH₃⁺) on PS to interact with the carboxylate groups on G_i3.

We have previously shown that negatively charged lipids, particularly PG, stabilize the β₂AR in an active-like conformation as revealed by changes in mB−β₂AR fluorescence and an increased affinity for agonists¹⁴. These effects are likely due to interactions between the lipids and positively charged amino acids on the β₂AR. Here we observed that the effect of DOPG and DOPS on mB-β₂AR can be reversed by both Ca²⁺ and Mg²⁺ (Fig. 5a). Yet, these divalent cations do not appear to reduce coupling to G_s.

β₂AR signals from caveolin-rich rafts^33,34 within T-tubules¹⁹. While β₂AR preferentially interacts with PG in insect cell membrane¹⁴, the phospholipid composition immediately adjacent to β₂AR in T-tubules, and how it changes during β₂AR trafficking, is currently unknown. Net-neutral PC and PE are the major phospholipids in T-tubules^35–37. However, negatively charged PS is enriched in T-tubules relative to other membrane fractions (7.5−12.3% of total phospholipid)^35–39. While cytosolic Ca²⁺ concentrations are typically less than 1 mM⁴⁰, concentrations of Ca²⁺ in the mM range may be observed in cardiac myocytes.

Investigators have long speculated about the functional role of Ca²⁺ in the cleft between the T-tubule membrane (where β₂AR is localized) and the juxtaposed sarcoplasmic reticulum (SR)^21,22,41. During each action potential, extracellular Ca²⁺ flows into the cleft through L-type Ca²⁺ channels (LTCCs) on the plasma membrane and through ryanodine receptors (RyRs) on the sarcoplasmic reticulum⁴⁰. Cleft Ca²⁺ concentrations spark to >100 μM in the absence of epinephrine and >1 mM^42,43 following epinephrine stimulation, a consequence of G_s activation. Computational models show that negatively charged phospholipids buffer approximately half the Ca²⁺ released into the cleft⁴², and experiments have shown that 80% of inner-leaflet bound Ca²⁺ is bound to negatively charged phospholipids⁴⁴. Additionally, biochemical investigations show that Ca²⁺ can cluster negatively charged PS⁴⁵ and PIP₂^46,47 lipids.

β₁AR and β₂AR signaling through G_s alters calcium handling in the cardiac myocyte, and increases the magnitude of Ca²⁺ currents and Ca²⁺ transients, which stimulate cardiac contraction^40,48. However, elevated Ca²⁺ concentrations also activate the Ca²⁺/calmodulin-dependent protein kinase II (CaMKII), which promotes apoptosis, and is implicated in structural remodeling that ultimately results in cardiac dysfunction^9,49–52. Elevated Ca²⁺ also activates calcineurin, which exacerbates pathological hypertrophy²¹. For these reasons, β₁AR-selective (and nonselective) beta blockers have proven to be efficacious medicines for treatment of heart failure⁵³. However, β₂AR, owing to its dual G_s/G_i selectivity, is functionally distinct from the strictly G_s-coupled β₁AR, and several lines of evidence suggest β₂AR−G_i signaling functions to keep G_s signaling in check via negative feedback: β₂AR−G_i signaling occurs minutes after β₁AR−G_s and β₂AR−G_s signaling², β₂AR−G_i signaling suppresses changes in calcium handling^48,54, and β₂AR−G_i signaling is antiapoptotic^7,8. While the mechanism that triggers β₂AR−G_i signaling is unknown, our biochemical observations suggest elevated Ca²⁺ concentrations could trigger β₂AR coupling to G_i3. It is notable that overexpression of the Ca²⁺/sodium exchanger facilitates β₂AR−G_i suppression of β₁AR−G_s signaling⁵⁵, and overexpression has been cited to increase the inward LTCC Ca²⁺ current⁵⁶.

It is also notable that intracellular Ca^{2+ 57,58} and Mg^{2+ 58} concentrations rise during ischemia and rise even higher during reperfusion. Whether the rising concentrations facilitate β₂AR-G_i signaling is unknown. However, β₂AR−G_i signaling can reduce the extent of cardiac necrosis caused by ischemia and reperfusion⁵⁹.

We were surprised that we did not observe mB−β₂AR coupling to G_i2 under conditions where we observed coupling to G_i1 and G_i3 (Fig. 1c, Supplementary Fig. 3). Although G_i2 contains the αN EDGE motif, we did not observe mB−β₂AR−G_i2 interaction in the presence of Ca²⁺ (Fig. 5d). While the sequences of Gα_i1 and Gα_i3 are 94% identical, Gα_i2 shares less sequence identity with Gα_i1 and Gα_i3, 88% and 86%, respectively. Supplementary Fig. 7 shows the location of the amino acid differences in G_i2 relative to both G_i1 and G_i3 using the recent cryo-electron microscopy structure of the adenosine A1-G_i2 complex (PDB: 6D9H⁶⁰). It can be seen that these amino acids do not appear to interact directly with the receptor in the nucleotide-free complex. However, these amino acids may form weak interactions with the poorly ordered intracellular loop 3 that is not observed in the structure, or may interact with other domains of the receptor during complex formation. Previous studies provide evidence for at least one transient intermediate state in formation of the β₂AR−G_s complex⁶¹. Both G_i2 and G_i3 couple to β₂AR in cardiac myocytes³. However, during heart failure, G_i2 expression is commonly upregulated^62–64. In contrast, G_i3 upregulation, while reported⁶⁵, is less commonly observed. How selective G_i2 upregulation influences β₂AR−G_i signaling is not fully understood. Signaling through G_i2 and G_i3 could play different roles in cardiac physiology. It is possible that β₂AR coupling to G_i3 plays a role in preventing myocyte damage during transient ischemia or prolonged periods of adrenergic stimulation, such as during exercise. Both of these conditions would be associated with elevated cytosolic Ca²⁺. In contrast, β₂AR coupling to G_i2 may play a more prominent role in the failing heart.

G proteins are a large superfamily, grouped into four subfamilies (G_s, G_i/o, G_q/11, G_12/13) encoded by 16 different genes⁶⁶. Each subfamily activates distinct signaling pathways, and functional effects are cell-type specific. Most GPCRs can signal through more than one G protein subfamily, and ongoing research attempts to identify mechanisms that regulate G protein selectivity within a cell⁶⁶. Whether local membrane charge affects G_i interaction with other G_i-coupled GPCRs is not currently known. In cardiac myocytes, β₂AR signals from a PS-enriched, Ca²⁺-enriched microenvironment, which highlights the potential relevance of our biochemical observations. Additionally, the observation that Ca²⁺ sensing receptor (CaSR) switches from G_q to G_i after cytosolic Ca²⁺ increases⁶⁷ is also potentially relevant to our findings. However, knowledge that Ca²⁺ facilitates G_i interaction with membrane is not sufficient to predict how Ca²⁺ might affect G_i interaction with other receptors.

In conclusion, we show that local membrane charge differentially modulates β₂AR interaction with competing G protein subtypes (G_s and G_i). This discovery expands our knowledge of mechanisms that regulate the G protein coupling selectivity of GPCRs.

Methods

G protein expression and purification

All G proteins were human heterotrimeric G proteins (Gα, Gβ₁, Gγ₂). The Gβ₁ subunit contained an N-terminal 6xHis tag followed by a rhinovirus 3C protease site used for purification⁶¹. Gα_s was the short splice variant. The G_i3-G_s chimera was created starting with Gα_s, replacing residues 1−38 (the αN helix) with the equivalent region of Gα_i3 (residues 1−31 of Gα_i3)⁶⁸. G_s-neg. was a G_s mutant with residues 32−35 of Gα_s (KDKQ) replaced with the equivalent region in Gα_i3 (EDGE). G_i3-pos. was a G_i3 mutant with residues 25−28 of Gα_i3 (EDGE) replaced with the equivalent region in Gα_s (KDKQ). G protein was expressed in Tni insect cells (Expression Systems Cat. 94-002S) using two recombinant baculoviruses, a virus encoding Gα and a separate virus encoding both the Gβ₁ and Gγ₂ subunits. Following infection, cells were incubated for 48 h at 27 °C, harvested by centrifugation, and suspended in lysis buffer (10 mM Tris (pH 7.5), 100 μM MgCl₂, 5 mM β-mercaptoethanol (BME), 10 μM GDP, and protease inhibitors). The membrane fraction was collected by centrifugation and solubilized using a Dounce homogenizer and buffer comprised of 20 mM HEPES (pH 7.4), 100 mM NaCl, 1% sodium cholate, 0.05% n-dodecyl-β-d-maltopyranoside (DDM), 5 mM MgCl₂, 5 mM BME, 10 μM GDP, and protease inhibitors. The soluble fraction was isolated by centrifugation, G protein was captured on Sepharose Fast Flow (GE Healthcare) charged with nickel, and gradually exchanged into SEC buffer (20 mM HEPES (pH 7.4), 100 mM NaCl, 0.05% DDM, 100 μM TCEP, 10 μM GDP, 1 mM MgCl₂). 3C protease was added to cleave G protein off the resin, and the G protein was dephosphorylated using calf intestinal alkaline phosphatase, antarctic phosphatase, and lambda protein phosphatase. Subsequently, G protein was isolated in SEC buffer using a Superdex 200 10/300 GL column (GE Healthcare). The main peak corresponding to heterotrimeric G protein was collected, concentrated, and frozen.

Receptor expression, purification, and labeling

The β₂AR construct was PN1¹⁴, where human WT β₂AR (R16, Q27 variant) is modified to contain an N-terminal FLAG tag, a C-terminal rho-1D4 tag, a TEV protease cleavage site between V24 and T25, and a 3C protease cleavage site between G365 and Y366. Additionally, mutations were introduced to increase expression (M96T, M98T), to remove a glycosylation site (N187E) and to remove reactive cysteines (C378A, C406A). β₂AR was expressed in Sf9 insect cells (Expression Systems Cat. 94-001S) using recombinant baculovirus and media supplemented with 1 μM alprenolol. Cells expressing β₂AR were harvested by centrifugation and suspended in lysis buffer (10 mM HEPES (pH 7.4), 1 mM EDTA, 1 μM alprenolol, and protease inhibitors). The membrane fraction was collected by centrifugation and solubilized using a Dounce homogenizer and buffer comprised of 20 mM HEPES (pH 7.4), 100 mM NaCl, 1% n-dodecyl-β-d-maltopyranoside (DDM), 0.03% CHS, 2 mM MgCl₂, 1 μM alprenolol, and protease inhibitors. The soluble fraction was isolated by centrifugation and anti-FLAG (ATCC HB-9259) affinity chromatography was used to purify β₂AR, remove alprenolol, and adjust detergent concentration (to 0.1% DDM, 0.01% CHS). Monobromobimane (mB, Thermo Fisher Scientific) labeling was then performed overnight with excess mB in the presence of 100 μM TCEP and the reaction was quenched with 5 mM l-cysteine before further purification. All β₂AR preparations were functionally purified by alprenolol-Sepharose affinity chromatography and washed on an anti-FLAG column to remove ligand. The eluted β₂AR was dialyzed in buffer comprised of 20 mM HEPES (pH 7.4), 100 mM NaCl, 0.1% DDM, and 0.02% CHS, concentrated, and dephosphorylated using lambda protein phosphatase. For experiments assessing phosphorylation, unphosphorylated β₂AR was phosphorylated with protein kinase A in the presence of 2 mM ATP. β₂AR was stored frozen. Phosphorylation was assessed using the Pro-Q Diamond Phosphoprotein Gel Stain (ThermoFisher Scientific), per the manufacturer’s instructions. Stained acrylamide gels were scanned with a Typhoon 9410 Imager (GE Healthcare). Recombinant human NTSR1 (residues 20−418) contained an A85^1.54L mutation to increase expression. NTSR1 was purified from Sf9 insect cells into buffer comprised of 20 mM HEPES (pH 7.5), 100 mM NaCl, 5% glycerol, 0.01% lauryl maltose neopentyl glycol (LMNG), 0.001% CHS, and 0.5 µM JMV 449, and stored frozen.

Ligands

(−) epinephrine was purchased from Sigma (purity > 99%). JMV 449 was purchased from Tocris (purity 97.6%).

Micelle composition

n-dodecyl-β-d-maltopyranoside (DDM), cholesteryl hemisuccinate (CHS), and 1-palmitoyl-2-oleoyl-sn-glycero-3-(PE,PC,PG,PS) lipids (Avanti Polar Lipids) were mixed in the indicated ratios and solubilized in chloroform. Chloroform was evaporated, and the films were re-suspended in 20 mM HEPES (pH 7.4), 100 mM NaCl.

Fluorescence spectroscopy

In experiments examining mB-β₂AR in micelles, mB−β₂AR was preincubated (30 min room temperature) in micelle stock prior to dilution with other reaction components. pH 7.4 HEPES buffer (containing 100 mM NaCl, ± ligand, ± CaCl₂ or MgCl₂) and G protein were sequentially included. Mixtures were incubated 2.5–3.0 h at room temperature. Final mB−β₂AR concentration was 100−300 nM. Emission spectra were read at 22 °C using a Fluorolog-3 spectrofluorometer (Horiba Jobin Yvon Inc.). (Bandpass = 4 nm; Excitation = 370 nm; Emission = 420−500 nm, collected in 1 nm increments). The raw S1c/R1c spectra were smoothed using Prism (GraphPad Software) (n = 15 neighbors, second-order polynomial). Lambda max is defined as the wavelength at which fluorescence emission is maximum. To determine the EC₅₀, data were fit to the agonist vs. response model in Prism 7.0d software.

GTP turnover

Where lipid environments were compared, samples were prepared as they were for fluorescence spectroscopy. Following β₂AR incubation with G protein, 1 μM GTP + 5 μM GDP mixtures (final concentration) were added to initiate the GTP turnover reaction. Reaction buffer contained 20 mM HEPES (pH 7.4) and 100 mM NaCl. Where G_i2 and G_i3 were compared, the GTP turnover reaction was initiated by mixing a solution containing 10 μM GTP and ligand-bound receptor (4 μM β₂AR + 800 μM epinephrine or 1 μM NTSR1 + 10 μM JMV 449) in equal volume with a solution containing 2.5 μM G protein, 20 μM GDP, 20 mM MgCl₂, and 200 μM TCEP. At the indicated timepoints, the GTP remaining was assessed using the GTPase-Glo assay (Promega), which detects GTP using a luminescence readout. Luminescence was detected using a SpectraMax Paradigm plate reader equipped with a TUNE SpectraMax detection cartridge (Molecular Devices). Background luminescence was subtracted from experimental reactions.

Statistics

Two-way and three-way ANOVA were performed using Graphpad Prism 7.0d.

Electrostatic modeling

Structural views and mutant models were generated using PyMOL (Schrödinger, LLC). We selected rotomer positions that most closely matched those seen in PDB 3SN6⁶⁹ (for the G_i3-pos. model) and PDB 1GP2⁷⁰ (for the G_s-neg. model). Continuum electrostatics models were calculated using the APBS⁷¹ plugin (MG Lerner, University of Michigan, Ann Arbor) for PyMOL. Atomic charge and radii were calculated using the online PDB2PQR server⁷² (pH 7.4, PARSE force field, hydrogen bond optimization, clash avoidance).

Nanodisc reagents

1,2-dioleoyl-sn-glycero-3-(PE,PC,PG,PS) lipids (Avanti Polar Lipids) were used because of their low phase transition temperature. Lipids were dissolved in buffer comprised of 20 mM HEPES (pH 7.5), 100 mM NaCl, 50 mM sodium cholate, 1 mM EDTA at 16.6 mM, and were sonicated on ice before use. The MSP belt was MSP1E3D1⁷³. pMSP1E3D1 was a gift from Stephen Sligar (Addgene plasmid #20066). The protein was expressed in BL21(DE3) E. coli (Cat. 70235-3 Millipore Sigma), and cells were lysed with sonication in 10 mM Tris-HCl, 100 mM NaH₂PO₄, 6 M GuHCl, 1% Triton X-100 (pH 8.0). The soluble fraction was isolated by centrifugation and passed through a Sepharose Fast Flow (GE Healthcare) column charged with nickel. Immobilized protein was washed with buffer comprised of 10 mM Tris-HCl, 100 mM NaH₂PO₄, 6 M GuHCl, and 0.2% Triton X-100 (pH 7.0), and then washed with buffer comprised of 50 mM NaH₂PO₄, 300 mM NaCl, and 0.2% Triton X-100 (pH 8.0). The protein was eluted using 250 mM imidazole. Impurities were precipitated by two rounds of heating at 70 °C for 1 h; after each round, the soluble fraction was isolate by centrifugation. The final soluble fraction containing MSP1E3D1 was supplemented with 20 mM sodium cholate and purified further using a HiPrep 16/60 Sephacryl S-300 HR size-exclusion column (GE Healthcare) equilibrated with buffer comprised of 20 mM HEPES, 100 mM NaCl, 1 mM EDTA, and 20 mM sodium cholate. The main peak corresponding to MSP1E3D1 was collected, dialyzed in 20 mM HEPES, 100 mM NaCl, 1 mM EDTA, 5 mM sodium cholate, concentrated to 728 μM, and frozen.

Nanodisc reconstitution

Nanodiscs were formed with one lipid type (i.e. 100% DOPS, DOPG, DOPE, or DOPC). Reconstitution was initiated by sequentially mixing water, 20× reconstitution buffer (400 mM HEPES (pH 7.5), 2 M NaCl, 20 mM EDTA), lipid, mB−β₂AR, and MSP1E3D1. The volume of mB−β₂AR stock added to the mixture was ~13% of the final mixture volume. MSP1E3D1 and mB−β₂AR were added 1:10 (molar ratio). Lipid and MSP1E3D1 were added 35:1 (molar ratio). The mixture was incubated at 4 °C for 2 h. Subsequently, Bio-Beads SM-2 resin (107.1 mg per μmol lipid, Bio-Rad) were added (4 °C for 4 h) to remove detergent, which triggers the reconstitution of mB−β₂AR in nanodisc bilayers. The soluble fraction was isolated by centrifugation. Bare nanodiscs were separated from nanodiscs containing mB−β₂AR using anti-FLAG (ATCC HB-9259) affinity chromatography. The eluate was incubated with 5 mM EDTA at 4 °C for ≥ 1.5 h to remove divalent cations. Subsequently, samples were injected into a Superdex 200 10/300GL size-exclusion column (GE Healthcare) equilibrated in buffer comprised of 20 mM HEPES (pH 7.4), 100 mM NaCl, and the main peak corresponding to nanodisc mB−β₂AR was harvested, concentrated, and frozen. The concentration of nanodisc mB−β₂AR was approximated by SDS-PAGE, using detergent solubilized β₂AR as a protein concentration reference standard.

Reporting summary

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

Author contributions

M.J.S. designed, performed, and interpreted the research, and wrote the manuscript. B.K.K. championed the investigation, advised on the project, and edited the manuscript. M.J.S. performed the β₂AR and G protein purifications, labeling, nanodisc reconstitutions, and all of the experiments, and S.M., D.H., M.M. and Y.D. provided valuable technical assistance as described: S.M. advised on cloning and provided G protein for pilot experiments. D.H. advised on G protein purification and the GTP turnover assay. M.M. advised on nanodisc reconstitution and provided MSP1E3D1 protein. Y.D. collaborated on pilot experiments not included.

Data availability

The source data underlying Figs. 1a, c−f, 2a, b, 3b, 4, 5a–d, and Supplementary Figs. 1a−c, 2a, b, 3a, b, 4, and 6 are provided as a Source Data File. pMSP1E3D1 is available from Addgene (#20066). All PDB files that were analyzed have been published before and can be obtained from the RCSB Protein Data Bank using the accession codes 5JQH, 3SN6, 1GP2, and 6D9H. A reporting summary for this Article is available as a Supplementary Information file. All other datasets supporting the findings of the study are available from the corresponding author on reasonable request.

Competing interests

B.K.K. is a co-founder of and consultant for ConfometRx, Inc. The other authors declare no competing interests.

Supplementary information

Supplementary Information accompanies this paper at 10.1038/s41467-019-10108-0.