A highly conserved δ-opioid receptor region determines RGS4 interaction

Abstract

The δ-opioid receptor (δ-OR) couples to Gi/Go proteins to modulate a variety of responses in the nervous system. Τhe regulator of G protein signalling 4 (RGS4) was previously shown to directly interact within the C-terminal region of δ-OR using its N-terminal domain to negatively modulate opioid receptor signalling. Herein, using molecular dynamics simulations and in vitro pull-down experiments we delimit this interaction to 12 helix 8 residues of δ-ΟR and to the first 17 N-terminal residues (NT) of RGS4. Monitoring the complex arrangement and stabilization between RGS4 and δ-OR by molecular dynamics simulations combined with mutagenesis studies, we defined that two critical interactions are formed: one between Phe329 of helix8 of δ-ΟR and Pro9 of the NT of RGS4 and the other a salt bridge between Glu323 of δ-ΟR and Lys17 of RGS4. Our observations allow drafting for the first time a structural model of a ternary complex including the δ-opioid receptor, a G protein and a RGS protein. Furthermore, the high degree of conservation among opioid receptors of the RGS4-binding region, points to a conserved interaction mode between opioid receptors and this important regulatory protein.

4Box

AbbreviationsRGS domain of RGS4

5-HT1A

serotonin 1A receptor

6xHis

hexahistidine

AMF

100 μm AlCl₃, 2 mΜ MgCl₂, 100 mM NaF, 1 mM GDP

CT

carboxyl terminus

DSLET

[D-Ser²]-Leucine-enkephalin-Thr⁶

ERK1,2

extracellular signal-regulated protein kinase

GAP

GTPase-activating protein

GPCR

G-protein-coupled receptors

GST

glutathione-S-transferase

GST

glutathione-S-transferase

HA

haemaglutinin

HRP

horseradish–peroxidise

MAPK

mitogen-activated protein kinase

NMS

normal mouse serum

NRS

normal rabbit serum

NT

N-terminus

ORs

opioid receptors

PAR

Protease-activated receptor

PMSF

phenylmethylsulfonyl fluoride

PVDF membrane

polyvinylidene difluoride membrane

RGS4ΔPL

RGS4 mutant that Pro9 and Lys17 residues are replaced with alanine

RGS

regulators of G protein signalling

STAT5B

signal transducer and activator of transcription 5B

δ-CT

δ-opioid receptor carboxyl terminal tail

δ-OR

δ-opioid receptor

ΔΝ₁₇RGS4

RGS4 lacking the first 17 amino acid residues of the N-terminus

ΔΝRGS4

RGS4 lacking its N-terminal domain

Introduction

The δ-opioid receptor (δ-ΟR) belongs to family A of rhodopsin-like G-protein-coupled receptors (GPCRs) and is expressed in the central and peripheral nervous system to mediate diverse physiological responses ranging from pain perception, anxiety, respiratory depression and neurotransmitter release to cell proliferation and neuronal differentiation [1, 2]. Τhe δ-OR mediates these responses via coupling to members of the Gi/o proteins [3, 4], however, ample experimental evidence has demonstrated that it also interacts with various accessory proteins, which alter the effectiveness of agonist-driven cell signalling, trafficking and cellular localization of this receptor [4]. Of interest is that the C-terminal domain (CT), and particularly helix 8, plays a key role for these interactions. Pull-down assays encompassing the CTs of δ- μ- and κ- opioid receptors, as well as functional assays in living cells, have previously demonstrated that RGS4, a member of the B/R4 family of regulators of G protein signalling (RGS) proteins, directly interacts with all three opioid receptor subtypes using its N-terminus (NT) domain to negatively regulate their signalling [5-7].

Regulator of G protein signalling proteins comprise a large and diverse family of proteins (> 30 family members) that directly bind to Gα subunits to attenuate their signalling [8]. Experimental evidence has demonstrated that RGS proteins can also interact with GPCRs, as well as with other additional partners, to serve receptor functions distinct from their classical G protein signalling [9-12]. RGS4 is expressed in developing neurons and functional studies have linked it to the regulation of opioid, cholinergic and serotonergic signalling in the brain [13]. RGS4 was found to negatively modulate signalling of various GPCRs including 5-HT1A [14] and δ-, μ- and κ-ORs [5-7] and to be involved in δ-ΟR-mediated behaviours [15]. Recent findings demonstrated that RGS4 is part of a multicomponent signalling complex encompassing the CT of δ-ΟR (δ-CT), specific members of the Gi/o protein(s) and STAT5B involved in δ-OR-mediated neurotropic events in neuronal cells [12, 13, 16].

The precise GPCR-G-RGS protein combinations determine the nature and duration of a given response. Little is known on how RGS selectivity for G proteins and GPCRs is determined in living cells [17, 18]. RGS proteins directly bind to the three switch regions of the Gα subunit through their RGS conserved helical domain and stabilize them in a transition state conformation [19]. The structural determinants of RGS-Gα protein interactions are being characterized [20, 21]. Functional assays combined with structure-based analyses have determined the structural features involved in the interaction between G proteins and of a large array of human RGS proteins [20], however, the molecular mechanisms dictating RGS selectivity have not been fully answered yet. For example, in the case of RGS2, recent combined energy calculations and GTPase activity measurements suggest that RGS-Gαq specificity is determined by three RGS2-specific residues of the RGS domain [22], whereas other findings reveal that the chief determinant of RGS8 for Gαq are most likely found in switch III of Gα [21]. On the other hand, up to date there is no clear indication on the way RGS proteins pair with GPCRs. We have previously shown that RGS4 interacts with all three opioid receptor subtypes (δ, μ, κ) in an agonist independent manner to confer selectivity for a particular subset of G protein(s) and to negatively modulate their signalling [6, 7, 12]. Other observations have shown that RGS4 and RGS2 are recruited to the plasma membrane by G proteins and/or the expressed receptors [8, 10]. Recent evidence indicates that RGS2 and RGS4 interact directly with PAR1 in a Gα-dependent manner to modulate PAR1/Gα-mediated signalling [23]. All these observations indicate that selectivity of RGS proteins for a given GPCR or G protein is influenced by different parameters ranging from the nature and abundance of G and RGS proteins present in a certain cellular milieu to the activation state of each receptor. Taken all these into account, the identification of a consensus site for RGS binding with receptors would help clarify how RGS proteins, receptors and G proteins interplay modulate signalling.

In the present work, we provide evidence of δ-ΟR-RGS4 direct association demonstrating the importance of the 17 residues of the initial part of the RGS4 NT and of 12 residues of the δ-CT (DENFKRCFRQLC) in helix 8. Our structural insights, combined with mutagenesis studies, modelling and molecular dynamics (MD) simulations, allowed us to identify critical interacting residues in the δ-CT and the NT of RGS4 and propose a ternary complex organization including RGS4, δ-OR and Gα_i1,3. This model could be useful towards delineating key binding residues determining RGS-GPCR coupling selectivity.

Results

Modeling RGS4-δ-OR interaction

It was previously demonstrated that RGS4 directly interacts with the three opioid receptor (OR) subtypes μ-, δ- and κ- using its N-terminal domain (NT) and that deletion of the first 57 amino acids of the NT of RGS4 abolished this interaction suggesting that this region is responsible for opioid receptor recognition and binding [5-7]. In order to clarify the molecular details of this interaction, we used various modelling approaches combined with molecular dynamics simulations. We mainly focused on the preferential conformational state of the NT of RGS4 under native conditions. Previous studies suggested that RGS4 binds to the membrane through an amphipathic α-helix [24]. To support this finding and to evaluate the possibility of an α-helical arrangement of the NT-RGS4, we modelled an amphipathic helix for residues M1 to F25 of the NT region based on the membrane anchor domain (1–31) of the nonstructural protein 5A of the hepatitis C virus (PDB 1R7E). Molecular dynamics simulations monitored the stability of the α-helical arrangement under different conditions (Fig. 1). Our data suggest an α-helical conformation of the NT region up to Leu23 when placed in a favourable amphipathic environment (Fig. 1A, 8 × 100 ns). Stability of this region as an α-helix was further confirmed by running it in an alternative membrane composition as well as starting from a de novo predicted model obtained using QUARK (Fig. 1B, 8 × 100 ns and C 8 × 100 ns). On the basis of these findings, we next explored the interaction of the helical NT-RGS4 with the δ-OR. For this reason, we started from the system which provided the best environment for NT-RGS4 helical stability (system 2, as seen in Fig. 1A) and included the crystallized helix 8 region (YAFLDENFKRCFR) of the δ-OR (PDB ID: 4EJ4). The choice of this receptor region was based on previous observations demonstrating that this domain is critical for δ-OR-RGS4 interaction [6]. Additionally, knowing that helix 8 is amphipathic and parallel to the membrane plane, in a similar way as the NT-RGS4, allowed us to hypothesize that this interaction mode could be the most plausible scenario. This is further supported by the fact that the two palmitoylation sites at the NT-RGS4 (C2 and C12) are known to be important for GPCR recognition, trafficking, protein targeting to the membrane and inhibition of G protein activation and signalling [6, 24, 25]. In order to appropriately sample other possible interaction modes, we followed the evolution of complex formation of di-palmitoylated NT-RGS4 and the helix 8 of δ-OR during 12 replicates of 100 ns (starting system in Fig. 2A). Monitoring complex rearrangement and stabilization, we found that two critical interactions are frequently formed in a stable arrangement of the NT-RGS4 and helix 8 of the δ-OR. The first being a stacking interaction between Phe329^8.54 (helix 8 of δ-OR) and Pro9 (NT-RGS4) and a second a stable salt bridge established between Glu323^8.48 (helix 8 of δ-OR) and Lys17 (NT-RGS4) (Fig. 2B).

Figure 1

Molecular dynamics (MD) simulations of the conformational stability of the N-terminal-RGS4 (NT-RGS4) region under different conditions. (A) To analyse the stability considering different NT-RGS4 membrane insertions in a DPPG:DPPC bilayer, in system 1 (left) the modelled amphipathic NT-RGS4 α-helix (M1 to F25) was placed at 22 A along the z direction allowing for only slight contacts with the membrane (4 × 100 ns); whereas in system 2 (right), NT-RGS4 was placed at 21 A along the z direction allowing for abundant contacts with the membrane lipids (4 × 100 ns). (B) To verify whether the membrane environment affects the stability of the model, system 2 from A using a DPPG: DPPC bilayer (left, 4 × 100 ns) was compared to an alternative system 2 (right, 4 × 100 ns), in which NT-RGS4 was placed in a POPC bilayer. (C) To evaluate the stability of an alternative ab initio model of the NT-RGS4 region, simulations of this second model were run in DPPG: DPPC (left, 4 × 100 ns) and POPC (right, 4 × 100 ns) bilayers. For each system, α-helical stability is indicated along residues M1 to F25 by grey bars.

Figure 2

Interaction between NT-RGS4 and helix 8 of δ-OR. (A) Molecular dynamics simulation of interaction between NT-RGS4 and helix 8 of δ-OR. (B) Top view of complex formation between NT-RGS4 and helix 8 showing critical interactions (i) stacking between Phe329^8.54 and Pro9 and electrostatic interaction between Glu323^8.54 and Lys17. For simplicity no palmitoylation is shown. (C) The NT1-17 of RGS4 is responsible for δ-OR interaction. Upper panel: Wild-type recombinant 6xHis-RGS4 and 6xHis-ΔΝ₁₇RGS4 (0.5 μm each) were incubated with the GST-δ-CT and pull downs were performed as described in Methods. Bound proteins were subjected to SDS/PAGE, transferred to PVDF and detected using an anti-His antibody (1 : 3000). Lower panel: The amount of GST fusions loaded was verified after stripping and reprobing with an anti-GST antibody (1 : 5000). Results shown are representative of three independent experiments.

A specific region at the NT-RGS4 is responsible for δ-OR binding

To validate the results obtained by MD simulations, suggesting that the region including the amino acids Pro9 and Lys17 of RGS4 are indeed critical for δ-ΟR association, we generated initially an RGS4 mutant lacking the first 17 amino acid residues of the NT (ΔΝ₁₇RGS4). As shown in Fig. 2C, pull-down experiments using a GST fusion peptide encompassing the δ-CT, together with the truncated ΔΝ₁₇RGS4 and subsequent immunoblotting with an anti-His antibody failed to retain RGS4 binding (lane 4). In contrast, a 24 kDa polypeptide band corresponding to wild-type RGS4 was detected in similar pull downs (lane 3). No bands corresponding to RGS4 or ΔΝ₁₇RGS4 were detected using GST alone (lanes 1 and 2). These results suggest that the initial 17 amino acid residues of RGS4 are responsible for binding with δ-OR, providing the first biochemical evidence for the importance of this region of RGS4 for δ-OR interaction.

Pro9 and Lys17 in RGS4 are critical for δ-ΟR association

In order to validate the importance of Pro9 and Lys17 residues of RGS4 for δ-OR association, we replaced them with alanine (RGS4ΔPL) and tested the ability of the RGS4ΔPL mutant to interact with δ-ΟR and the δ-CΤ. For that reason, the δ-CT was incubated with HEK293 cell lysates expressing the wild-type HA-RGS4 and the HA-RGS4ΔPL. As shown in Fig. 3A, incubation with wild-type RGS4 and subsequent immunoblotting with an anti-HA antibody revealed a polypeptide corresponding to RGS4 (lane 1) as verified by a band corresponding to RGS4 in the cell lysate (lane 3). In contrast, in lysates expressing the RGS4ΔPL mutant, only a very faint binding was detected (lane 2), suggesting that RGS4ΔPL loses its ability to interact tightly with the δ-CT.

Figure 3

RGS4 amino acid residues Pro9 and Lys17 are responsible for δ-OR interaction. (A) HEK293 cells were transiently transfected with 3 μg of flag-δ-ΟR and 3 μg of each wild-type RGS4 and its mutant. 300 μg of HA-RGS4 and HA-RGS4ΔPL expressing HEK293 cells lysates were incubated with approximately 1 μm of the GST fusion encompassing the δ-CT for 20 min at 4 °C and pull-down experiments were performed as described in Methods. Protein complexes were separated on 12% SDS/PAGE and bound wild-type RGS4 and RGS4ΔPL was detected with an anti-HA antibody (1 : 1000). The amount of GST fusion loaded was verified after stripping and reprobing with an anti-GST antibody (1 : 5000). Results shown are representative of three independent experiments. (B) HEK293 cells were transfected with 5 μg of HA-RGS4 or HA-RGS4ΔPL and flag-δ-ΟR and were challenged or not with 1 μm DSLET for 5 min. Immunoprecipitation of the cell lysates using an anti-HA antibody (1 : 1000) revealed the presence of co-precipitated flag-δ-ΟR (upper panel). Mock transfected cells were used as control (lane 1). (C) Effect of RGS4 and the RGS4ΔPL mutant on δ-OR-driven ERK1,2 phosphorylation. HEK293 cells stably expressing the flag-δ-OR were transiently transfected with empty vector, HA-RGS4 or HA-RGS4ΔPL. 48 h post-transfection, cells were stimulated with 1 μm DSLET for 5 min and cell lysates were resolved in 12% SDS/PAGE. The phosphorylated ERK1,2 was visualized by immunoblotting with an antiphospho-ERK 1,2 antibody (1 : 1000) (upper panel). Equal loading was verified after stripping and reprobing the PVDF membranes with tubulin (1 : 1000) (lower panel). Results shown are representative of three independent experiments.

To further verify the significance of Pro9 and Lys17 residues of RGS4 for δ-ΟR interaction, lysates from HEK293 cells transiently expressing the flag-δ-OR together with the wild-type HA-RGS4 or HA-RGS4ΔPL, were immunoprecipitated with anti-HA and immunoblotted with an anti-flag antibody. As demonstrated in Fig. 3B, RGS4 co-immunoprecipitated with the δ-ΟR as identified by the same molecular weight band in cell lysates (compare lanes 2 with 5). This band was absent in lysates expressing the RGS4ΔPL (lane 3) and in mock transfected cells (lane 1). Additionally, the activation of the δ-OR with DSLET did not alter binding of the RGS4ΔPL with the receptor (lane 4), suggesting that RGS4ΔPL mutant loses its ability to interact with δ-OR even in its activated state in HEK293 cells.

It was previously shown that RGS4 has a negative effect on δ-OR-mediated ERK1,2 phosphorylation [6]. To examine whether RGS4ΔPL expression alters ERK1,2 phosphorylation in response to δ-ΟR activation, HEK293 cells expressing the δ-OR were challenged with DSLET. Αs shown in Fig. 3C, 1 μm DSLET enhanced ERK1,2 phosphorylation after 5 min stimulation of δ-OR. This phosphorylation as expected was abolished in the presence of wild-type RGS4 (compare lanes 3 with 4). However, when similar measurements of MAP kinase phosphorylation were performed in the same cells, expressing RGS4ΔPL, there was a significant activation of ERK1,2 in the presence of DSLET (compare lanes 5, 6). These results suggest that RGS4ΔPL loses its functionality and thus does not interfere in δ-OR-mediated ERK signalling, confirming the importance of Pro9 and Lys17 residues of RGS4 for δ-ΟR interaction.

Helix 8 of δ-OR determines RGS4 interaction

The C-terminal region (CT) of ORs shares a conserved domain forming helix 8 known to be critical for Gα, Gβγ subunit and RGS4 binding [5, 6, 26]. Based on our molecular dynamics results, and in an attempt to define the importance of the proximal element of helix 8 of the δ-CT we examined whether a synthetic peptide DENFKRCFRQLC (i4), formed by residues 322–333 of the δ-OR (Fig. 4A), is critical for RGS4 binding. Gel electrophoresis under nondenaturating conditions using the i4 peptide preincubated with purified recombinant 6xHis-RGS4 after immunoblotting indicated an upward shift in the mobility of RGS4 (Fig. 4B, lanes 5–8) compared with that in its absence (lane 3). At high concentrations (300 and 500 μm) of i4 peptide, the RGS4-peptide complex migrates at even higher molecular weight due to charge difference (lanes 7, 8), suggesting that peptide i4 forms a complex with RGS4 that migrates at a slower rate. In contrast, when similar experiments were performed using the truncated ΔΝ₁₇RGS4 preincubated with increasing concentrations of i4 peptide, no shift in protein migration was observed, suggesting that the ΔΝ₁₇RGS4 charge and its hydrodynamic size remained unchanged (Fig. 4B, lanes 14–16). Similarly, no alterations in protein charge were detected using the control peptide MELVPSAR (Fig. 4B, lanes 9 and 10). These data suggest that deletion of the seventeen NT amino acid stretch of RGS4 prevents binding to the i4 peptide, demonstrating for the first time that helix 8 residues 322 to 333 of the δ-OR and the NT17 residues of RGS4 constitute the interface between these proteins.

Figure 4

RGS4 forms a complex with the i4 peptide. (A) Schematic diagram of the δ-OR. The amino acid sequence of i4 peptide is shown in cyan. (B) 62 μm of the purified His-RGS4 or ΔΝ₁₇RGS4 were incubated with increasing concentrations (10–500 μm) of i4 peptide in the presence of DTT and electrophoresed in a 10% PAGE under nondenaturating conditions as described in Methods. Proteins were detected using an anti-His antibody (1 : 3000) after immunoblotting.

Structural features of the RGS4-Giα-δ-OR complex

We have previously demonstrated that RGS4 forms selective complexes with specific Gα subunits upon δ-OR activation in HEK293 cells and that RGS4 and Gα form a heterotrimeric complex within amino acids 311–336 of the δ-CT [5, 6]. To explore the nature of this ternary complex, we tested whether the ΔΝ₁₇RGS4, which does not interact with the δ-CT, can form a stable heterotrimeric complex with prebound active Gαt and the δ-CT. To do that we initially characterized RGS4 and Gα binding by preincubating purified recombinant 6xHis-RGS4 and/or ΔΝ₁₇RGS4 with Gαt_GDP in the presence or absence of AlF⁻⁴ and Mg²⁺ (AMF). As shown in Fig. 5A, RGS4, as well as the ΔΝ₁₇RGS4 mutant, interact with Gα both in the presence or absence of AMF. To further explore whether ΔΝ₁₇RGS4 coupled to Gα has the capacity to form a ternary complex with the δ-OR, similar pull-down experiments were performed in the presence of δ-CT. As shown in Fig. 5B, only wild-type RGS4 forms a ternary complex with Gαt and the δ-CT in the presence of AMF (lane 6, upper, middle and lower panels). ΔΝ₁₇RGS4 was not detected upon preincubation with Gαt_GDP in the presence of AMF, despite Gα binding to the δ-CT (lane 5, upper panel). As anticipated, no bands corresponding to ΔΝ₁₇RGS4 were detected in the absence of Gat at any conditions tested (Fig. 5B, middle panel lanes 3 and 4). These results suggest that deletion of the initial 17 amino acid of the NT of RGS4 abolishes heterotrimeric complex formation between RGS4, active Gαt and the δ-CT.

Figure 5

ΔΝ₁₇RGS4 does not complex with Gαi and the δ-CT. (A) A fixed amount of 6xHis-RGS4 and ΔΝ₁₇RGS4 (0.5 μm of each) bound to Ni-NTA agarose beads were incubated with purified Gαt_GDP (0.5 μm) in PBS with protease inhibitors and 1 mm GDP in the presence or absence of AMF for 10 min at room temperature and 20 min at 4 °C. Bound proteins were subjected to SDS/PAGE, transferred to PVDF and detected using an anti-Gα (1 : 1000) and His antibodies (1 : 3000). Results shown are representative of three independent experiments. (B) A fixed amount of GST-δ-CT (1 μm) was bound to glutathione-sepharose beads as described in Methods. Purified Gαt_GDP, ΔΝ₁₇RGS4, or RGS4 (0.5 μm each) were incubated with PBS, protease inhibitors in the presence or absence of AMF for 20 min at 4 °C. The δ-CT bound to glutathione-sepharose beads was added to the above samples and further incubated under rotation for 20 min at 4 °C. Bound proteins were subjected to SDS/PAGE, transferred to PVDF membranes and detected using anti-Gα, anti-His and anti-GST antibodies. Upper and middle panels represent bound proteins visualized after immunoblotting with Gα and anti-His antibodies. Lower panel represents equal protein loading verified after stripping and reprobing using an anti-GST antibody (1 : 5000).

Discussion

Up to date there is no clear indication on the way RGS proteins interact with GPCRs and which is the spatial organization for this interaction. Our challenge was thus to define the critical residues involved in δ-ΟR-RGS4 association. Our MD simulations suggested that the 12 residues of δ-OR helix 8 and the first 17 NT residues of RGS4 could constitute an interface for these two proteins. They also pointed Pro9 and Lys17 located at the NT of RGS4 and Phe329^8.54 and Glu323^8.48 located in helix 8 of the δ-CT as important contact points between these proteins. Further validation by in vitro pull-down experiments indicated that a truncated RGS4 version lacking the first 17 NT residues (ΔΝ₁₇RGS4) retained Gα coupling but prevented RGS4-δ-OR-Gα ternary complex formation. This suggests that the initial 17 residues of NT-RGS4 are required for RGS4 coupling with δ-ΟR and Gα. Furthermore, using a decapeptide encompassing part of helix 8, indicated that although wild-type RGS4 interacted within this domain, ΔΝ₁₇RGS4 was unable to interact. These results suggest that the interaction of δ-OR with RGS4 can be narrowed down to the domain DENFKRCFRXXC (322–333) forming helix 8 of the δ-OR. Notably, this region is highly conserved among all three (δ, μ and κ) opioid receptor subtypes (Fig. 6A).

Figure 6

Proposed ternary complex of the δ-OR, RGS4 and Gα. (A) Alignment of helix 8 of the δ, κ and μ-opioid receptors. (B) Diagrammatic presentation of the possible scenarios of δ-ΟR-Giα and δ-ΟR-RGS4-Giα interaction. (C) Interaction of RGS4 and Gαi through the RGS-box would still allow for the NT of RGS4 to reach helix 8 of the δ-OR establishing the interactions detected by our computational and experimental approaches. (PDB ID: 1AGR, orange transparent) and its modelled NT (pink transparent), the active δ-OR (cyan transparent) and the crystallized active Gαi (PDB ID: 2OED, grey transparent).

Our next challenge was to validate the importance of Pro9 and Lys17 residues of RGS4 for δ-OR association. Knowing that RGS4 has an antagonistic effect on effectors, we examined whether the RGS4ΔPL expression in HEK293 cells modulates δ-OR signalling. Opioid receptors stimulate ERK1,2 activity via Gi/o proteins [27] and RGS4 is implicated in δ-ΟR-mediated MAP kinase phosphorylation [6, 7]. In this study we demonstrate that although RGS4 presence attenuates MAP kinase phosphorylation, this inhibitory effect was fully reversed in cells expressing the mutant RGS4ΔPL, suggesting that Pro9 and Lys17 point mutations of RGS4 have a functional role in δ-ΟR-mediated signalling. On the other hand, co-immunoprecipitation studies indicated that RGS4ΔPL does not interact with δ-ΟR in living cells either constitutively or upon δ-ΟR activation, suggesting that mutations of Pro9 and Lys17 impair RGS4 association with the δ-OR. Additionally, in vitro pull downs using the δ-CT and cell lysates expressing RGS4ΔPL indicated a profound decrease in RGS4 binding to the δ-CT but not complete abolition. This residual RGS4ΔPL binding could be attributed either to other residues within helix 8 retaining a weak RGS4 binding, or another plausible scenario could be that RGS4ΔPL couples with the δ-CT within its RGS domain of RGS4 (4Box). The latter explanation is supported by previous findings demonstrating that the 4Box is capable to bind to the C-terminus of μ-opioid receptor [5, 6].

Collectively, the present work allows us to propose the following organization for the three interacting proteins (Fig. 6B). In the absence of RGS4, Gα is able to interact with the δ-CT region irrespective of its activation state as was also previously shown [6]. However, when RGS4 is present, this protein could outcompete Gα-δ-CT interaction. In this way, RGS4 binding to the δ-ΟR through its N-terminal region and to Gα through its RGS box could result in a transient ternary complex that recruits activated Gα to the proximity of the receptor and then favours Gα inactivation through the GAP function of RGS4. Although further experiments will be needed to determine the timeline at which these interactions occur during the Gα activation cycle, our observations indicate that the interfaces of RGS4 with the Gα and δ-ΟR do not overlap and consequently allow us to propose the formation of a transient ternary complex presented in Fig. 6B. Taking all these into account and combining information on the interactions delimited by our experiments with recent structural evidence on different opioid receptor activation states [28] and previous Gαi-RGS4 crystallized complexes [29], we propose a first schematic 3D arrangement of δ-OR, RGS4 and Gα complex (Fig. 6C).

The selectivity of coupling of a given GPCR to specific G proteins is critical for exerting a physiological response. Likewise, the identification of a recognition site of an RGS protein with a given receptor will allow understanding receptor-binding selectivity determinants and their impact on specific molecular responses. Opioid analgesia and tolerance development involve complex cellular and molecular mechanisms. RGS4 has been found to modulate Gi protein signalling and accelerate the early rate of δ-OR endocytosis, whereas truncation of the N-terminal region of RGS4 was unable to regulate the internalization pathway of δ-OR [6]. RGS proteins have been found to be involved in phenomena of addiction and tolerance and addictive drugs control the expression levels of several RGS proteins [30]. The mRNA levels of RGS4 are the highest in the brain compared to other RGS proteins and RGS4 has been shown to play key role in synaptic signalling and plasticity and to be involved in many brain diseases [13] and δ-ΟR-mediated behaviours [15]. Indeed, abnormal RGS4 function has been implicated in schizophrenia [31, 32] anxiety [33], Parkinson [34] and Alzheimer's disease [35], suggesting that RGS4 is a multifunctional protein with several roles and functions required for many cellular responses. There has been a lot of interest in the development of specific RGS-inhibitors as targets to regulate G protein signalling [36, 37]. Importantly, our results suggest that interaction between opioid receptors and RGS4 occurs through a highly conserved receptor region, an observation that could point to a common interaction mechanism. In this sense, the present data can provide information for the development of small molecules that can target such complexes to regulate cell physiology, as well as to control the duration of the action of opioids and prevent the adverse effects of tolerance and dependence.

Methods

Computational methods: Part 1

The N-terminal tail (NT) of RGS4 adopts an amphipathic α-helical conformation in a membrane composition of DPPC : DPPG 80 : 20 [24]. In order to validate this possible conformational arrangement, we first modelled the N-terminal tail of RGS4 based on the solved amphipathic membrane anchor domain of the nonstructural protein 5A (NS5A) of hepatitis C virus (PDB ID 1R7E) using the moe software (https://www.chemcomp.com/). This template was selected out of 8 in-plane membrane (IPM) anchors with 3D coordinates found in the Protein Data Bank: 1H0A, 1HUR, 2R45, 4FTB, 1Q4G, 3NT1, 2SQC, 1R7E. An initial sequence alignment of the amphipathic region to the NT of RGS4 using the MOE modelling package (https://www.chemcomp.com/) yielded an extremely low sequence identity with < 20%. In addition to sequence identity, we used as second criteria the number of positions that conserve any of the amino acid properties: hydrophobic (Ala, Val, Ile, Leu, Met, Trp, Phe, Tyr), hydrophilic (Ser, Thr, Asn, Gln, Glu, Asp, Arg, Lys, His), acidic (Glu, Asp), basic (Arg, Lys, His), aromatic (Phe, Tyr, Trp, His), aliphatic (Val, Leu, Ile, Gly, Ala, Met, Pro) and polar (Asn, Ser, Thr, Gln). Using these criteria the best template was 1R7E with a sequence identity of 17.2% and 17 residues of conserved properties. The modelled NT of RGS4 was inserted into an amphipathic environment using the CHARMM GUI membrane builder (http://www.charmm-gui.org/) using a membrane composition of DPPC : DPPG 80 : 20. Two different starting systems probing two different placements of NT-RGS4 with respect to the membrane plane were built and then subjected to NPT equilibration for 20 ns (Fig. 1A). As a control strategy, and taken into account previous studies demonstrating the helical character of the NT region in membranes which are not conventionally used for MD simulation [24], we run the RGS4 NT region in a POPC bilayer further demonstrating that it retains its helical features irrespective of the membrane system used in simulations (Fig. 1B). Given the low homology of the NT of RGS4 to any available structurally solved template, we also generated a de novo protein structure prediction using quark [38] as an additional control strategy. All the five resulting models presented helical structures, with one of them being compatible with cysteine palmitoylation in residues 2 and 12 resulting in membrane anchoring of the helix. Simulations of this alternative model further confirmed the helicity and amphipathic character of this region in our simulations (Fig. 1C). For all systems, we run four replicates of 100 ns in NVT ensemble yielding a total simulation time of 0.8 μs.

Part 2

To assess the potential interaction between the modelled NT of RGS4 (obtained in part 1) with the crystallized helix 8 of the δ-opioid receptor, we implemented both proteins into an amphipathic membrane environment using the CHARMM GUI membrane Builder (membrane composition DPPC : DPPG 80 : 20). Three different starting systems were built and then subjected to NPT equilibration for 20 ns. Afterwards, we run for each starting system six replicates of 100 ns in NVT ensemble yielding a total simulation time of 1.8 μs. The above described procedure was applied to test the preferred interaction of the unmodified NT of RGS4 (system 1: 1.8 μs) as well as the post-translational modified NT of RGS4 (system 2: palmitoylated Cys2 and Cys12 – 1.8 μs) with helix 8.

Simulation conditions for part 1 and part 2

All simulations were carried out using the simulation software acemd [39]. In a first step, the systems were equilibrated using the NPT ensemble with a target pressure equal to 1.01325 bar, a time-step of 2 fs and using the RATTLE algorithm for the hydrogen atoms. In this stage, the harmonic constraints applied to the heavy atoms of the protein and ligand were progressively reduced from an initial value of 10 kcal·mol⁻¹·Å⁻¹ until an elastic constant force equal to 0 kcal·mol⁻¹ and the temperature was increased to 300 K. The purpose of this relaxation phase is to allow a complete adjustment of membrane lipids to the receptor, thus filling nonphysiological gaps between protein and membrane lipids. All the simulations were conducted using the same nonbonded interaction parameters, with a cut off of 9 Å, a smooth switching function of 7.5 Å and the nonbonded pair list set to 9 Å. For the long range electrostatics, we used the PME methodology with a grid spacing of 1 Å. In a third stage, production phases were performed using the NVT ensemble with aforementioned parameters but a time-step of 4 fs and a hydrogen scaling factor of 4. This time step is possible due to the implementation of the hydrogen mass repartitioning scheme in the ACEMD code. Importantly, individually generated starting structures allow a more robust statistical analysis and thus the detection of relevant dynamic events that are independent from the starting structures.

Part 3: Building the complete complex comprising RGS4, δ-opioid receptor and Gα_i

The active structure of the δ-opioid receptor was initially modelled based on two templates using the moe software (standard settings). Thereby, the crystallized μ-opioid receptor (6DDE) was used as template for the receptor core, whereas helix 8 was modelled based on the active β-adrenergic receptor (3SN6).Then, various steps of superimposition were carried out followed by manual adjustments:

• The previously modelled active structure of the δ-opioid receptor was superimposed to the crystallized μ-opioid receptor (6DDE).
• The RGS4 box (1AGR) was superimposed to crystallized RGS8 in complex with active Gi_alpha3 in 2ODE.

Finally, the modelled NT of RGS4 was connected to the RGS4 box and the whole system was minimized.

Constructs and reagents

Purified Gtα_GDP from bovine retina and the cDNA for 6xHis-tagged RGS4 were kindly provided by H. Hamm, Vanderbilt University, Nashville, TN and T. M. Wilkie, University of Texas, TX, USA respectively. Haemaglutinin (HA)-tagged human RGS4 cDNA in pcDNA3 was kindly provided by G. Milligan, University of Glasgow, Scotland. Proteases inhibitor cocktail was purchased from Roche (Roche Diagnostics, Penzberg, Germany), ECL western blotting substrate were purchased from Thermo Scientific (Waltham, MA, USA). Glutathione sepharose 4B beads and Protino Ni-NTA agarose beads were from Macherey & Nagel (Duren, Germany). Antibodies against GST and HA was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and His antibody was obtained from BD Pharmingen (San Jose, CA, USA). Mouse secondary antibody conjugated with HRP were purchased from KPL (Gaithersburg, MD, USA).

Generation of RGS4 mutants

Construction of a seventeen N-terminal amino acid deletion mutant of RGS4 (ΔΝ₁₇RGS4)

For the generation of the N-terminal deletion of rat RGS4, two oligonucleotides were constructed so that the 5′ end primer contained an NcoI site and a methionine (ATG) for translation initiation and the 3′ end primer an XhoI site: 5′-ATGCCCATGGATATGAAACATCGGCTG-3′ (forward) and 5′-ATACCTCGAGGGCACACTGAGGGACTAG-3′ (reverse). The PCR product was digested and cloned into the pET28b vector (Novagen, EMD Biosciences, Burlington, MA, USA) using T4 ligase. Positive clones were selected and the presence of the insert was verified by nucleotide sequence analysis. For protein expression 0.5 L of LB medium was inoculated and protein expression was carried out as described by Ref. [6].

Construction of Pro9 and Lys17 mutations of RGS4 (RGS4ΔPL)

For the generation of the double mutant of RGS4, amino-acids P9A and K17A were replaced to alanine using as a template the haemaglutinin (HA)-tagged RGS4 in pcDNA3 according to the Manufacturer's instructions as described in Q5 site directed mutagenesis kit (New England Biolabs, Ipswich, MA, USA). The primers used were as follows: 5′-GAGGAGTGCAGCAGATATGAAACATCGGCTAGG-3′ (forward) and 5′-AAGCAAGAAGCCGCCAGACCTGCAAGCCCTTT-3′ (reverse). Positive clones were selected and purified using the Midi-prep kit (Qiagen, Hilden, Germany) and verified by sequencing of the complete protein-coding region.

Cell cultures and transient transfections

HEK293 cells stably expressing the flag-δ-opioid receptor (δ-ΗΕΚ293) were grown in Dulbecco's modified Eagle's medium containing 2 mm glutamine, 100 U·mL⁻¹ penicillin, 100 μg·mL⁻¹ streptomycin and 10% fetal bovine serum under 5% CO₂ at 37 °C. Transient transfections of wild-type RGS4 (WTRGS4) and the double-mutated RGS4 (RGS4ΔPL) were performed using the TurboFect in vitro transfection reagent (Thermo Scientific) according to the Manufacturer's instructions.

GST pull-down assays

Approximately 1 μm of the GST-fusion peptides encompassing the δ-CT were immobilized on glutathione sepharose 4B beads using PBS, pH 7.4 at 4 °C, containing protease inhibitor cocktail, 0.2 mm phenylmethylsulfonyl fluoride (PMSF), 20 μg·mL⁻¹ leupeptin and 20 μg·mL⁻¹ antipain. The mixture was washed three times with PBS and subsequently was incubated with purified recombinant 6xHis-tagged RGS4 and ΔΝ₁₇RGS4 for 15 min or with protein lysates from transiently transfected HEK293 cells expressing the wild-type RGS4 and the RGS4ΔPL mutant for 15 min following the procedure as described by Ref. [5].

Detection of MAPK phosphorylation

HEK293 cells stably expressing the δ-OR were transiently transfected with the empty vector pcDNA3, wild-type RGS4 and the double mutant of RGS4 (RGS4ΔPL) and cultured in six well plates for 48 h. Sixteen hours before the addition of drugs, the culture medium was removed and replaced by fresh serum-free medium. Agonists were added to the cells and allowed to incubate at 37 °C. Measurement of MAPK phosphorylation was performed as described by Ref. [40].

Peptide synthesis

Peptides were synthesized by an Applied Biosystems peptide synthesizer (model 430 A, Norwalk, CT, USA) as described by Ref. [26]. These were designated peptide i4, residues 322–333, DENFKRCFRQLC, derived from the C-terminal and peptide NH₂, residues 1–8, MELVPSAR encompassing the N-terminus of δ-OR.

Western blotting

Protein samples were resolved on SDS/PAGE (14%) and transferred to PVDF membranes as described by Ref. [39]. For the native gel electrophoresis, a fixed amount of 10 μg purified recombinant 6xHis-RGS4 and 6xHis-ΔΝ₁₇RGS4 were incubated with increasing concentrations of i4 and NH₂ peptides in the presence of 1 mm DTT for 2 h at 4 °C. Protein samples were subjected on a nondenaturating gel consisting of 10% bis-acrylamide, 40 mm acetic acid and 80 mm beta-alanine pH 4.3. The gel was incubated for 1 h to 1% SDS buffer and transferred to PVDF membranes as described above. Detection was performed using the enhanced chemiluminescence (ECL; Pierce-Thermo Scientific) and a luminescent image analyser (Fujifilm LAS-4000, Tokyo, Japan).

Acknowledgements

This work is supported by the GSRT grant ‘Excellence’ NO-ALGOS-3722 to ZG. All authors were members of the European COST Action CM1207 (GLISTEN), CK was supported by COST Action CM1207 through an STSM. JS and ZG are members of the COST Action CA18133 (ERNEST). JS acknowledges financial support from Instituto de Salud Carlos III FEDER (PI15/00460 and PI18/00094) and the ERA-NET NEURON & Ministry of Economy, Industry and Competitiveness (AC18/00030), whereas TMS from the National Science Centre of Poland, project number 2017/27/N/NZ2/02571. We thank Sofia Koutloglou NCSR ‘D’ for the construction of the RGS4ΔPL mutant and Leonidas J. Leontiadis for his contibution regarding the native gel electophoresis. The facilities of the Hellenic Research Infrastructure OPENSCREEN-GR were used for protein expression. We also acknowledge support by the ‘OPENSCREEN-GR’ ‘An Open-Access Research Infrastructure of Chemical Biology and Target-Based Screening Technologies for Human and Animal Health, Agriculture and the Environment’ (MIS 5002691) which is implemented under the Action ‘Reinforcement of the Research and Innovation Infrastructure’, funded by the Operational Programme ‘Competitiveness, Entrepreneurship and Innovation’ (NSRF 2014-2020) and co-financed by Greece and the European Union (European Regional Development Fund).

Conflict of interest

The authors declare no conflict of interest.

Author contributions

CK: designed, performed experiments, analysed data and contributed to manuscript writing; MM-S conducted the computational studies, analysed data and contributed to manuscript writing; AS: performed experiments; TMS conducted computational studies; ZG and JS: conceived the study, designed, analysed the data and contributed to manuscript writing.