Blog

By Weimin Conrad Hong

Excerpt from the article published in Journal of Pharmacology and Experimental Therapeutics, May 2020,  373 (2) 290-301; DOI: https://doi.org/10.1124/jpet.119.262790

Editor’s Highlights

• The sigma-1 receptor (σ₁R) interacts with and modulates the activity of a plethora of partner proteins, including channels, receptors, and transporters, and regulates their function.
• Changes in σ₁R oligomerization by ligands can modulate the availability of σ₁R to associate with its partners, but do not fully account for ligands’ efficacies in functional assays
• Multiple domains coordinate the oligomerization of σ₁R.
• N-terminal (NT) domain of σ₁R plays a crucial role in its multimerization by potentially linking two homotrimers to form a hexamer or multiple homotrimers to high-order oligomers.
• Intact NT interactions are required to initiate σ₁R high-order assembly.
• The equilibrium between monomers, dimers, and multimers of σ₁R is dynamically regulated by agonists or antagonists in distinct manners.
• Antagonists promote σ₁R multimerization, whereas agonists facilitate its dissociation. 
• Distinct regulation of σ₁R multimerization by agonists and antagonists may selectively modulate activities of clientele proteins within their interactome network. 
• σ₁R mutants implicated in neurodegenerative diseases displayed aberrant multimerization, suggesting that balance in σ₁R oligomerization is important in its physiologic function.
• σ₁R oligomerization may be precisely controlled by cells’ adaptive responses to physiochemical changes in the environment, which in turn may impact these signaling mechanisms.

Abstract

Extensive studies have shown that the σ₁ receptor (σ₁R) interacts with and modulates the activity of multiple proteins with important biological functions. Recent crystal structures of σ₁R as a homotrimer differ from a dimer-tetramer model postulated earlier. It remains inconclusive whether ligand binding regulates σ₁R oligomerization. Here, novel nondenaturing gel methods and mutational analysis were used to examine σ₁R oligomerization. In transfected cells, σ₁R exhibited as multimers, dimers, and monomers. Overall, σ₁R agonists decreased, whereas σ₁R antagonists increased σ₁R multimers, suggesting that agonists and antagonists differentially affect the stability of σ₁R multimers. Endogenous σ₁R in rat liver membranes also showed similar regulation of oligomerization as in cells. Mutations at key residues lining the trimerization interface (Arg119, Asp195, Phe191, Trp136, and Gly91) abolished multimerization without disrupting dimerization. Intriguingly, truncation of the N terminus reduced σ₁R to apparent monomer. These results demonstrate that multiple domains play crucial roles in coordinating high-order quaternary organization of σ₁R. The E102Q σ₁R mutant implicated in juvenile amyotrophic lateral sclerosis formed dimers only, suggesting that dysregulation of σ₁R multimeric assembly may impair its function. Interestingly, oligomerization of σ₁R was pH-dependent and correlated with changes in [³H](+)-pentazocine binding affinity and B_max. Combined with mutational analysis, it is reasoned that σ₁R multimers possess high-affinity and high-capacity [³H](+)-pentazocine binding, whereas monomers likely lack binding. These results suggest that σ₁R may exist in interconvertible oligomeric states in a dynamic equilibrium. Further exploration of ligand-regulated σ₁R multimerization may provide novel approaches to modulate the function of σ₁R and its interacting proteins.

SIGNIFICANCE STATEMENT The σ₁ receptor (σ₁R) modulates the activities of various partner proteins. Recently, crystal structures of σ₁R were elucidated as homotrimers. This study used novel nondenaturing gel methods to examine σ₁R oligomerization in transfected cells and rat liver membranes. Overall, agonist binding decreased, whereas antagonist binding increased σ₁R multimers, which comprised trimers and larger units. σ₁R multimers were shown to bind [³H](+)-pentazocine with high affinity and high capacity. Furthermore, mutational analysis revealed a crucial role of its N-terminal domain in σ₁R multimerization.

Introduction

The σ receptor was named after the distinct behavioral signs induced by SKF10047 (N-allylnormetazocine) in a chronic spinal dog preparation (Martin et al., 1976). However, molecular cloning identified a 25-kDa membrane protein as the σ₁ receptor (σ₁R) (Hanner et al., 1996; Jbilo et al., 1997). Its sequence is highly conserved in evolution but distinct from opioid receptors, as originally proposed. Multiple alternative splice variants of σ₁R have been characterized (Pan et al., 2017), including an isoform lacking exon 3 (ΔE3), which encodes amino acids (aa) 119–149 (Ganapathy et al., 1999).

Extensive studies have shown that σ₁R can interact with and modulate the activity of a plethora of partner proteins, including channels, receptors, and transporters (Hayashi and Su, 2001, 2007; Aydar et al., 2002; Wu and Bowen, 2008; Carnally et al., 2010; Kim et al., 2010; Navarro et al., 2010; Balasuriya et al., 2012; Kourrich et al., 2013; Srivats et al., 2016; Hong et al., 2017; Sambo et al., 2017; Thomas et al., 2017) that play important roles in cellular homeostasis and neuronal signaling. The majority of σ₁R protein is located in endoplasmic reticulum, particularly mitochondria-associated endoplasmic reticulum membranes (Hayashi and Su, 2007), which are critical sites to modulate energy balance, calcium regulation, and stress response. Several σ₁R mutations have recently been implicated in juvenile amyotrophic lateral sclerosis and distal hereditary motor neuropathies. Molecular mechanisms underlying such motor neuron deficits have been studied intensively, and aberrant σ₁R expression and function appear to be crucial in these conditions (Al-Saif et al., 2011; Bernard-Marissal et al., 2015; Li et al., 2015; Gregianin et al., 2016; Watanabe et al., 2016; Dreser et al., 2017).

Many clinical drugs and synthetic compounds with diverse structures exhibit varying affinities for σ₁R (Matsumoto, 2007; Cobos et al., 2008; Maurice and Su, 2009; Chu and Ruoho, 2016) and appear to share a limited pharmacophore consensus (Walker et al., 1990; Ablordeppey and Glennon, 2007; Newman and Coop, 2007; Weber and Wunsch, 2017). Several candidate endogenous ligands have been proposed over the years, including neurosteroids (Su et al., 1988; Bergeron et al., 1996), sphingosine (Ramachandran et al., 2009), and N,N-dimethyltryptamine (Fontanilla et al., 2009), but these hypotheses have not been conclusively confirmed. Traditionally σ₁R ligands have been classified as agonists or antagonists, depending upon whether they produce or block certain cellular, physiologic, or behavioral responses. Although affinities of these ligands for σ₁R have been extensively studied using traditional binding techniques, molecular mechanisms for agonists or antagonists to induce distinct changes of σ₁R remain largely unknown.

The ability to modulate σ₁R function with different ligands has made it an attractive target for developing novel therapeutic strategies. It has been shown that σ₁R agonists have ameliorative effects in several animal models of neurodegenerative disorders, such as Alzheimer’s disease (Lahmy et al., 2013; Maurice and Goguadze, 2017; Ryskamp et al., 2019), Parkinson’s disease (Francardo et al., 2014), Huntington’s disease (Ryskamp et al., 2017), and retinal degeneration (Wang et al., 2016), whereas σ₁R antagonists have pain-relief effects (Merlos et al., 2017). Accumulating evidence also suggests that σ receptors are critically involved in cellular adaptive mechanisms elicited by psychostimulants (Cai et al., 2017; Katz et al., 2017) and alcohol (Sabino and Cottone, 2017). Therapeutic potentials of σ₁R antagonists have been explored in rodent models of cocaine or methamphetamine addiction (Hiranita et al., 2011; Robson et al., 2014; Sambo et al., 2017).

Whether σ₁R possesses one or two transmembrane domains (TMs) has been controversial (Hanner et al., 1996; Aydar et al., 2002; Hayashi and Su, 2007). In recent years, atomic force microscopy and solution NMR methods were employed to explore σ₁R structures (Carnally et al., 2010; Balasuriya et al., 2012; Ortega-Roldan et al., 2013). Breakthrough on the crystal structures of σ₁R has elucidated its homotrimer organization, with each protomer containing a single TM and a cytoplasmic ligand-binding pocket (Schmidt et al., 2016). Such structural architecture differs from a dimer-tetramer model postulated by early work (Chu and Ruoho, 2016). Furthermore, crystal structures of σ₁R bound with agonist (+)-pentazocine or antagonist haloperidol showed similar homotrimer organization with limited conformational rearrangement (Schmidt et al., 2018), suggesting that trimers may be the lowest free-energy state of σ₁R during crystallization. It remains unclear whether ligand binding affects the native high-order organization of σ₁R. This study examined σ₁R oligomerization using molecular, biochemical, and pharmacological techniques. The results show that multiple domains on σ₁R coordinate its multimerization. Furthermore, agonists and antagonists dynamically regulate σ₁R oligomerization in distinct manners, and quaternary structures of σ₁R significantly impact ligand binding.

…

Results

Previously, we reported a nondenaturing gel method that detected σ₁R monomer, dimer, or high molecular mass (MM) multimers based on their apparent electrophoretic mobility. The proportion of σ₁R multimers was decreased by the agonist (+)-pentazocine but increased by the antagonist CM304 (Hong et al., 2017). In this assay, during gel electrophoresis, SDS was replaced by PFO, a mild detergent shown to preserve protein oligomers (Ramjeesingh et al., 1999; Penna et al., 2008).

After literature review, σ₁R drugs that are well-characterized as agonists or antagonists were selected for this study. HEK293 cells were stably transfected with the WT human σ₁R containing N-terminal FLAG and 2xHis₈ tags (FH-σ₁R) (predicted MM of 32 kDa: 25 kDa σ₁R + 7 kDa tags with linker) and cultured to confluency in multiwell plates to minimize samples variability. Cells then were incubated with drugs in culture medium at 37°C for 1 hour, washed, and solubilized using a mild detergent, GDN. Lysates were run in PFO-PAGE and immunoblotted with FLAG antibodies. The two lower-MM bands matched the estimated size of σ₁R monomer (comprising one σ₁R polypeptide or protomer) and dimer (comprising two protomers). High-MM diffused bands likely represented multiple forms apparently larger than trimers (three promoters), and are thus termed as “multimers” in this study. Oligomerization of σ₁R is defined as assembly from monomers to any form containing at least two σ₁R protomers, whereas multimerization specifically refers to formation of σ₁R complexes containing three and more protomers.

σ₁R agonists (+)-pentazocine, DTG, (+)-SKF10047, and cocaine all significantly decreased σ₁R multimer band density, with (+)-pentazocine producing largest effect (32% ± 4% of vehicle, Fig. 1A). In contrast, σ₁R antagonists (BD1063, BD1047, BD1008, and haloperidol) all significantly increased the band density of σ₁R multimers above 2-fold of that in vehicle (Fig. 1B), resembling effects of CM304. Because they mostly have nanomolar affinities for σ₁R, low micromolar concentrations of these drugs were likely sufficient to permeate through cell membranes and occupy most of intracellular σ₁R sites during incubation and induce significant effects on its multimerization.

Distinct effects of (+)-pentazocine were dose-dependent, ranging from 0.1 to 2 μM (Fig. 1C). Further, pre-exposure of antagonists (haloperidol, BD1008, or BD1063) in these cells for 0.5 hours blocked (+)-pentazocine’s effects (Fig. 1, D and E). Notably, these drugs did not change the total pool of σ₁R (shown in SDS-PAGE) but altered the proportion of multimers to dimers and monomers. Dose-dependent effects of haloperidol (0.1–1 μM) on reversing (+)-pentazocine’s effects were also shown in PFO-PAGE (Yano et al., 2018). These data showed that generally σ₁R agonists and antagonists induced opposite effects on σ₁R multimerization. Effects by PRE-084 and NE-100 were consistent as agonist or antagonist, respectively, although these were not statistically significant.

Other mild detergents were then explored. If lysates were mixed with 2% sodium deoxycholate, only high-MM σ₁R multimeric bands were seen (unpublished data). Fortuitously, replacing 4% PFO with 2% SLS (Reichel, 2012), an ionic detergent less stringent than SDS, yielded remarkable results. σ₁R mainly showed as monomer and multimer bands, with the dimer band largely absent. Consistent with results in PFO-PAGE, (+)-pentazocine decreased, whereas haloperidol increased σ₁R multimers in SLS-PAGE (Fig. 2A). However, this assay had a higher sensitivity. Haloperidol increased σ₁R multimer band densities to approximately 4-fold of vehicle, and (+)-pentazocine decreased σ₁R multimer bands to 23% ± 4% of vehicle. Furthermore, in SLS-PAGE, significant effects on σ₁R multimeric band densities by NE-100 (157% ± 19% of control) and PRE-084 (67% ± 7% of control, Fig. 2B) were revealed. Interestingly, σ₁R multimers appeared as multiple high-MM smeared bands but were apparently larger than 100 kDa, suggesting that transfected σ₁R might exist in multimeric forms larger than homotrimers.

Because most σ₁R is located on intracellular membranes, during incubation, drugs permeated through cell membranes to bind σ₁R, and most likely they remained bound to σ₁R during cell lysis since (+)-pentazocine and haloperidol were shown to dissociate very slowly from σ₁R (dissociation t_1/2 > 3 hours) using traditional radioligand off-rate method (Bowen et al., 1993) or scintillation proximity assay (Schmidt et al., 2018). Furthermore, if lysates from drug-naïve cells were incubated with drugs overnight on ice, similar effects as those in preincubation were seen (Supplemental Fig. 1). Lastly, FH-σ₁R cell lysates exhibited robust binding of [³H](+)-pentazocine (Fig. 7), suggesting that GDN solubilization, to a large extent, preserved active conformations of σ₁R capable of ligand binding.

These features facilitated examination of effects by potential endogenous ligands of σ₁R. Several candidates have been proposed, including progesterone (Su et al., 1988), dehydroepiandrosterone (Bergeron et al., 1996), and d-erythro-sphingosine (Ramachandran et al., 2009). Because of their limited water solubility, it was difficult to incubate cells at concentrations close to their binding affinities for σ₁R (high nanomolars to low micromolars). However, these lipids could be dissolved in ethanol at 10 mM and then mixed with GDN-solubilized FH-σ₁R cell lysates to achieve final concentrations of 10–100 μM. After overnight incubation on ice, lysates were then subjected to SLS-PAGE. Compared with vehicle treatment (0.5% ethanol), progesterone (10 and 50 μM) significantly increased σ₁R multimers (144% ± 19% and 207% ± 26% of vehicle, Fig. 3). Dehydroepiandrosterone and d-erythro-sphingosine appeared to induce a slight, dose-dependent decrease of σ₁R multimers, albeit not statistically significant. Another proposed endogenous ligand for σ₁R, N,N-dimethyltryptamine (Fontanilla et al., 2009), as a Schedule I controlled substance, is not available at the current institution and was not tested.

In PFO-PAGE and SLS-PAGE, GDN lysates typically were mixed with PFO or SLS and heated at 37°C for 10 minutes before gel analysis. Distinct effects of agonists and antagonists suggest that they differentially affect the stability of σ₁R multimers. To test this idea, GDN lysates from drug-treated FH-σ₁R cells were mixed with PFO or SLS loading buffer and incubated at four different temperatures (25, 37, 50, and 70°C) for 10 minutes before being analyzed in PFO-PAGE or SLS-PAGE. The proportion of σ₁R multimer was gradually decreased by rising temperatures before disappearing at 70°C. Compared with vehicle, at each condition, (+)-pentazocine consistently decreased, whereas haloperidol increased σ₁R multimers (Fig. 4). Even at 50°C, haloperidol clearly protected σ₁R multimers. Hence, the antagonist haloperidol appeared to enhance the thermostability of σ₁R multimers, whereas agonist (+)-pentazocine had opposite effects.

Several approaches were used to allay the concern that epitope-tagged σ₁Rs in transfected cells may have different quaternary organizations than native σ₁R. First, effects of different epitope tags in stably transfected HEK293 cells were compared. Distinct effects by (+)-pentazocine and BD1008 were preserved for HA-tagged σ₁R (Supplemental Fig. 2A) in PFO-PAGE. Second, HEK293 cells were transiently transfected to express FH-σ₁R at different levels. Regardless of high or low expression, BD1008 increased, whereas (+)-pentazocine decreased FH-σ₁R multimers in SLS-PAGE and PFO-PAGE (Supplemental Fig. 2B).

Furthermore, whether drugs affect endogenous σ₁R multimerization was examined in rat liver tissues, where σ₁R is enriched (McCann and Su, 1991). Rat liver membranes were incubated with drugs and solubilized with GDN lysis buffer. Lysates were then subjected to nondenaturing gel analysis, and immunoblotted with a mouse monoclonal antibody for σ₁R (clone B5; Santa Cruz Biotechnology). Different from cell lysates, in PFO-PAGE σ₁R were detected as almost exclusively high-MM smear bands, very faint signals of dimer bands based on apparent electrophoretic mobility, and absence of monomers. Compared with vehicle, (+)-pentazocine and BD1008 appeared to increase or decrease the dimer signal, respectively (Fig. 5A). In SDS-PAGE, σ₁R in rat liver lysates only showed as a single band of 25 kDa if lysates were mixed with SDS (final 1%) and heated to 85°C. If the mixture was incubated at room temperature (RT) for 1 hour, a faint band near 75 kDa appeared in BD1008-treated samples (Fig. 5B). The apparent MM of this band was consistent with a σ₁R trimer, suggesting that in liver membranes, BD1008-induced σ₁R trimers were stable enough to partially resist SDS treatment at RT. Absence of nonspecific bands validated this antibody for detecting native σ₁R.

Most importantly, drug effects were convincingly shown when samples were run in SLS-PAGE (Fig. 5C). Compared with vehicle, (+)-pentazocine significantly decreased the proportion of multimers in total σ₁R proteins, with a concomitant increase in σ₁R monomers. In contrast, BD1008 had a significant effect opposite to (+)-pentazocine (Fig. 5, C and D). These drug effects were very similar to those seen in transfected HEK293 cells (Fig. 2), but the difference in σ₁R multimer MM was worth noting. In rat liver lysates, these bands appeared to range from approximately 70 kDa (possible trimer) to beyond 400 kDa, with the highest density near approximately 100 kDa, but those from cells appeared to have larger MM sizes (Fig. 2). This suggests that high-order quaternary organization of native or heterologously overexpressed σ₁R might not be the same, but drug effects on σ₁R multimerization were preserved overall. It should be cautioned that MM estimation of these bands was limited because of the nature of these gels.

Crosslinking assays were done to validate the presence of σ₁R multimers. DSP is a bifunctional crosslinker that selectively reacts with primary amines with a cleavable disulfide bond in its 12-Å spacer arm. DSP crosslinking in rat liver GDN lysates induced a 50-kDa band after nonreducing SDS-PAGE, which disappeared with tris(2-carboxyethyl)phosphine treatment to reduce disulfide bonds (Supplemental Fig. 3A). Because rat σ₁R has a sole lysine (Lys142), with its side chain solvent-accessible, based on human σ₁R crystal structures (Schmidt et al., 2016), the 50-kDa band was most likely a dimer formed through crosslinking at Lys142. FH-σ₁R has three more lysine residues in its epitope and linker. DSP crosslinking induced multiple high-MM bands. Beside a clear dimer band, two discernible bands were detected at positions corresponding to trimer and tetramer as well as smeared bands above 150 kDa (Supplemental Fig. 3B). These data suggest that in transfected cells, trimers of σ₁R could exist, yet they might undergo further assembly; thus, trimer bands were not detected in nondenaturing gels (Figs. 1 and 2).

Recent breakthrough on crystal structures of σ₁R shed new light on key residues mediating homotrimer formation (Schmidt et al., 2016). Mutants of σ₁R at these pivotal positions were examined in nondenaturing gels. In WT σ₁R, the benzyl side chains of Phe191 in three protomers form aromatic interaction with each other (Fig. 6A). Mutation of Phe191 to Gly (F191G) removed this interaction and abolished σ₁R multimerization but appeared to retain dimerization (lane 1, Fig. 6B). The hydrophobic side chain of Trp136 in WT σ₁R interacts extensively with several residues in the neighboring protomer. Substitution of Trp136 with Gly (W136G) showed severely reduced multimerization and a dimer but no monomer band (lane 2, Fig. 6B). Crystal structures show that Arg119 and His116 of a WT protomer form a network of hydrogen bonds with Asp195 and Thr198 of its neighboring protomers, which is critical in maintaining trimerization interface. Alanine substitution at either position (R119A or D195A) abolished multimerization but preserved dimerization, as observed in PFO-PAGE (lane 6 and 7, Fig. 6B). A σ₁R splice variant skipping exon3 (ΔE3, encoding aa 119–149) showed only as dimer but not multimer in PFO-PAGE (lane 3, Fig. 6B), confirming the essential roles of Arg119 and Trp136 in protomer multimerization.

Genetic studies have identified several σ₁R mutants that are implicated in neurodegenerative diseases with motor neuron deficits. The Glu102Gln (E102Q) mutant is associated with juvenile amyotrophic lateral sclerosis (Al-Saif et al., 2011). In contrast with WT σ₁R, it failed to form multimers but appeared as dimer exclusively in PFO-PAGE (lane 5, Fig. 6B). The WT σ₁R showed signals of strong monomer, weak multimer, and very faint dimer in SLS-PAGE, but the E102Q mutant and ΔE3 variant exhibited strong dimer signals, suggesting that such sequence alterations abolished σ₁R multimerization but promoted dimer formation (lane 3 and 5, Fig. 6C). Another disease mutant, Glu138Gln (E138Q) (Gregianin et al., 2016), also appeared to have impaired multimerization (unpublished data).

A recent study reported an important role of a GXXXG motif (residues 87–91) in oligomerization of σ₁R. Mutations replacing glycine with residues containing bulky aliphatic side chains appeared to abolish σ₁R multimers but preserve dimers and monomers, as examined in size-exclusion chromatography using σ₁R mutants expressed in Escherichia coli (Gromek et al., 2014). Two such mutants, Gly91Ile (G91I) and Gly87Leu-Gly88Leu (G87-88L), had low expression levels in transfected cells, as similarly observed in bacterial expression. The G91I mutant showed only as dimer, without multimer or monomer, in PFO-PAGE (lane 9, Fig. 6B). Weak signals of G87-88L were detected in SLS-PAGE, whereas in PFO-PAGE, little if no signals were seen. Overall, these results confirmed the importance of key residues at trimerization interface and the GXXXG motif in σ₁R multimerization.

Further mutational analyses were conducted to explore the dimeric interface. Very weak expression of double mutants (D195A/G91I, D195A/G87-88L, E102Q/G91I, and E102Q/G87-88L) prevented their signal detection in PFO-PAGE, despite faint bands in SLS-PAGE (unpublished data). Surprisingly, removing the N-terminal 36 residues (Δ36aa, including the TM region of aa 8–32) of σ₁R resulted in a mutant that showed as a distinct monomer in PFO-PAGE and SLS-PAGE (lane 11, Fig. 6, B and C). The mutant lacking N-terminal 10 residues (Δ10aa) displayed severely impaired oligomerization (lane 16). However, a mutant lacking first five residues (Δ5aa) but sparing the TM substantially retained multimerization (lane 15). Moreover, the E102Q mutant with Δ5aa remained as dimer (lane 14), but was converted to apparent monomer if combined with Δ36aa (lane 13). Similarly, Δ36aa deletion also changed dimeric G91I to monomer (lane 12).

These data suggest that in addition to key residues in the trimerization interface and GXXXG motif, the N-terminal (NT) domain of σ₁R plays a crucial role in its multimerization by potentially linking two homotrimers to form a hexamer or multiple homotrimers to high-order oligomers. In fact, the unit cell organization of σ₁R crystals shows a pair of homotrimers linked together through interactions of two parallel NT domains, each from a protomer of the two neighboring homotrimers (Schmidt et al., 2016). Further evidence was obtained in mutational analysis on σ₁R homomeric interaction by cotransfection of Myc-tagged WT σ₁R and a series of deletion mutants of σ₁R with N-terminal glutathione S-transferase fusion. Even upon gradual truncations of more than 100 residues in the C terminus of glutathione S-transferase–σ₁R, Myc-σ₁R was coenriched by glutathione beads pull-down. However, this interaction was substantially diminished if the NT of σ₁R was deleted (Supplemental Fig. 4).

Next, the impact of σ₁R oligomerization on ligand binding was studied. GDN-solubilized lysates from FH-σ₁R cells showed robust [³H](+)-pentazocine binding, with a K_d value (37.8 ± 3.3 nM, avg. ± S.E.M.) in a similar range to that of σ₁R expressed in Sf9 cell membranes (Schmidt et al., 2016) and a B_max value (20.3 ± 2.3 pmol/mg protein) several-fold higher than those in native tissues (McCann and Su, 1991; Bowen et al., 1993). Remarkably, σ₁R binding exhibited exquisite sensitivity to pH values in the buffer. Compared with normal buffer of pH 7.5, total binding (in the absence of unlabeled ligands) was markedly enhanced in a basic buffer of pH 9 but reduced in an acidic buffer of pH 6 (Fig. 7A). Kinetic analysis of these homologous competition-binding data (Fig. 7B) revealed that in pH 9, there was a significant decrease in K_d value (i.e., increase in affinity) of (+)-pentazocine rather than an increase in binding B_max (Fig. 7, C and D). In contrast, acidic buffer (pH 6) significantly decreased not only binding affinity but also B_max values of [³H](+)-pentazocine. These changes in ligand binding appeared to correlate with the oligomeric states of σ₁R revealed in PFO-PAGE. Compared with control (pH 7.5), there was a significant increase in σ₁R multimers at pH 9, with a concomitant decrease in dimers and monomers (Fig. 7E). An opposite effect was observed at pH 6. These data suggest that ligand binding to σ₁R is significantly affected by its quaternary structures.

Considering multiple oligomeric states of σ₁R, ideally [³H](+)-pentazocine binding should be analyzed using a multistate model. However, computer-assisted nonlinear regression of binding data showed that a simple, one-site model would suffice (Fig. 7, A and B), and Scatchard plot appeared to be linear (Fig. 7A inset). Statistical comparisons of one-site versus two-site models did not consistently yield clear-cut conclusions.

Reduced capacity to bind [³H](+)-pentazocine were observed in GDN lysates of σ₁R mutants deficient in multimerization, including ΔE3, E102Q, R119A, and D195A (Table 1). B_max values measured in untransfected HEK293 cells were negligible (2% of WT σ₁R-transfected cells, in pmol/mg protein), presumably because of a low level of endogenous σ₁R. The weak expression levels of these mutants in transfected cells confounded the interpretation of their low B_max. For instance, D195A σ₁R still showed approximately 20% of binding B_max of WT σ₁R despite its low expression, suggesting that (+)-pentazocine binding was not fully compromised in this dimer-forming mutant. Nevertheless, the monomer-only mutant Δ36aa σ₁R did not bind (+)-pentazocine, despite its sufficient expression. Together with observations that pH sensitivity of (+)-pentazocine binding correlated with changes of σ₁R multimerization (Fig. 7), these data support the hypothesis that σ₁R multimers exhibit most active conformation for high-affinity (+)-pentazocine binding, whereas its monomers likely do not bind (+)-pentazocine.

	σ1R
	WT	Δ36aa	R119A	D195A	E102Q	ΔE3
Bmax (pmol/mg protein)	16.6 ± 3.5	0.08 ± 0.05	1.83 ± 0.13	3.67 ± 0.26	0.99 ± 0.06	0.21 ± 0.22
Kd (nM)	16.4 ± 7.5	14.0 ± 3.2	9.8 ± 1.4	12.9 ± 2.7	5.0 ± 0.6	17.0 ± 14.1

Discussion

This study used two nondenaturing gel methods to examine σ₁R oligomerization. In general, agonists decreased σ₁R multimers, whereas antagonists increased multimers (Figs. 1–5). Antagonist binding appeared to stabilize σ₁R multimers, because a higher temperature was required to dissociate σ₁R multimers (Fig. 4).

Although these methods detected multiple high-MM smear bands, they could not determine the stoichiometry of σ₁R multimers. Detergent solubilization might introduce artificial aggregates of σ₁R. However, distinct changes in band signals by ligands argue against this. Furthermore, existence of high-MM σ₁R complex was supported by early purification studies using [³H]azido-DTG or [³H](+)SKF10047 as affinity ligands, in which the labeled protein complex under nondenaturing conditions appeared to be approximately 150 or 450 kDa (Kavanaugh et al., 1988; McCann and Su, 1991). In blue native gels, purified σ₁R showed as multiple smear bands from 60 to 480 kDa (Schmidt et al., 2016).

In rat liver membranes very little dimer and no monomer of σ₁R were present in PFO-PAGE (Fig. 5A), suggesting that native σ₁R multimeric complex was more resistant to extraction by PFO. Distinct drug effects on σ₁R multimerization were optimally demonstrated by SLS-PAGE (Fig. 5C). σ₁R multimers appeared as high-MM bands, including possible trimer, tetramer, and beyond. All σ₁R multimers disappeared in denaturing SDS-PAGE. A weak band of σ₁R trimer was induced by BD1008, if samples were not heated with SDS (Fig. 5B).

Notwithstanding their limitations, these nondenaturing gel approaches offered a relatively straightforward readout of σ₁R oligomerization. Results on G91I mutant were consistent between PFO-PAGE (Fig. 6B) and size-exclusion chromatography (Gromek et al., 2014). Two recent studies used Förster or bioluminescence resonance energy transfer assays to examine ligand effects on σ₁R oligomerization (Mishra et al., 2015; Yano et al., 2018). These assays are advantageous in monitoring the distance between donor- and acceptor-tagged σ₁R in live cells, whereas PFO-PAGE and SLS-PAGE appear to be more sensitive in detecting effects by some ligands, such as BD1047 and DTG.

To induce maximal effects on σ₁R multimerization, most drugs were used at low micromolar concentrations, approximately 100–1000-fold of their K_i values for σ₁R. (+)-Pentazocine showed dose-dependent effects on decreasing σ₁R multimers (Fig. 1C). With a subnanomolar affinity for σ₁R (James et al., 2012), CM304 induced significant effects at 0.1 μM. Although 0.45 μM drugs was sufficient to stabilize oligomers of σ₁R purified from bacteria (Gromek et al., 2014), 100 μM (+)-pentazocine or haloperidol was used in COS-7 cells for Förster resonance energy transfer analysis (Mishra et al., 2015). Thus, higher concentrations of drugs were necessary to permeate across membranes to bind intracellular σ₁R in cells. Most ligands tested here are relatively selective for σ₁R, but haloperidol also has a high affinity for dopamine D₂ receptors. This action was unlikely involved, because haloperidol had effects on σ₁R multimerization in cold cell lysates (Supplemental Fig. 1).

It is worth noting that effects of PRE-084 and NE-100 on σ₁R multimerization did not correlate with their binding affinities. The binding pocket in σ₁R can accommodate diverse ligands with a charged nitrogen as the central pharmacophore (Ablordeppey and Glennon, 2007). Comparison between σ₁R structures bound with (+)-pentazocine and NE-100 revealed limited conformational rearrangement (Schmidt et al., 2018). These structures likely provided snapshots of σ₁R in its lowest free-energy state. However, dynamic conformational changes induced by different ligands and impact on σ₁R quaternary structures have not been thoroughly examined. Further, drug binding to σ₁R may exhibit different cooperativity. Although (+)-pentazocine fully occupied the binding pocket of each σ₁R protomer in crystal structures, binding kinetic analysis of FLAG-tagged σ₁R and molecular dynamics simulation supported a multistep model of (+)-pentazocine binding (Schmidt et al., 2018). Whether PRE-084 or NE-100 induces different conformational changes or binding cooperativity of σ₁R from (+)-pentazocine will require more sophisticated techniques.

Stabilization of σ₁R multimer by antagonists may be explained by their preferential higher affinity for σ₁R multimers rather than dimers/monomers. In WT σ₁R haloperidol’s IC₅₀ value for competing against [³H](+)-pentazocine binding was approximately one-tenth of those in R119A or D195A mutants (unpublished data), which form only dimers and monomers. However, similar K_d values (Table 1) were not sufficient to explain why (+)-pentazocine dissociated σ₁R multimers. Other potential mechanisms, such as negative protomer cooperativity for (+)-pentazocine binding, will be pursued in future studies.

As σ₁R interacts with many protein partners and regulates their function (Schmidt and Kruse, 2019), efficacies of σ₁R drugs in functional assays are likely determined by multiple factors at molecular, cellular, and higher integrative levels. (+)-Pentazocine decreased σ₁R’s association with binding immunoglobin protein (Hayashi and Su, 2007), whereas haloperidol decreased its association with acid-sensing ion channels (Carnally et al., 2010), suggesting that different partners may preferentially interact with specific oligomeric forms of σ₁R. Changes in σ₁R oligomerization by ligands can modulate the availability of σ₁R to associate with its partners, but do not fully account for ligands’ efficacies in functional assays.

Data in this study suggest that multiple domains coordinate σ₁R oligomerization. Crystal structures of σ₁R have pinpointed critical residues mediating interactions for homotrimerization (Schmidt et al., 2016). Mutation of Arg119 or Asp195 abolished multimers in nondenaturing gels, confirming a crucial role of hydrogen bonds at these positions. Furthermore, hydrophobic interactions and van der Waals forces involving Phe191 and Trp136 are also pivotal, because removing their aromatic side chains severely impaired multimerization (Fig. 6, B and C). The importance of the GXXXG motif (Gromek et al., 2014) was also substantiated by results of G91I σ₁R in PFO-PAGE. Because Gly91 is in close proximity (4 to 5 Å) to Trp136 of a neighboring protomer in σ₁R crystal structures, mutation to bulky side chain disrupted multimerization.

In nondenaturing gels, multiple high-MM smeared bands appeared larger than σ₁R homotrimers, suggesting possible high-order organization of σ₁R trimers. A corollary of this scenario would require additional domains to mediate noncovalent assembly of trimeric building blocks. Current data support the hypothesis that the NT of σ₁R serves such a role. Unlike WT σ₁R and the Δ5aa mutant, Δ10aa and Δ36aa mutants yielded only monomers in PFO-PAGE (Fig. 6B) despite their intact trimerization domains. It is tempting to speculate that intact NT interactions are required to initiate σ₁R high-order assembly.

This provocative hypothesis is partly corroborated by σ₁R crystal structures, in which each unit cell comprises two pairs of homotrimers (Schmidt et al., 2016). Both pairs are linked together through interactions of two parallel NT domains, each from a protomer in the two homotrimers (Fig. 6A). The opposite orientation of the two pairs requires their embedding into two lipid bilayers in native membranes and is possibly an artifact in crystallization. However, dimerization of homotrimers may resemble a form of native σ₁R oligomerization.

Dimer formation after dissociation of multimers may involve assembly of two protomers from neighboring homotrimers via their NT interactions or rearrangement of two cognate protomers within a homotrimer. Current results on NT truncations favor the first scenario but do not rule out the latter. Dimer interactions appeared to be preserved by PFO but disrupted by the more stringent detergent, SLS. Persistence of E102Q or ΔE3 σ₁R dimers in SLS-PAGE is perplexing and requires further investigation. The E102Q mutant was shown to be prone to aggregation and have aberrant cellular location (Wong et al., 2016; Dreser et al., 2017). In coimmunoprecipitation assays, ΔE3 σ₁R exhibited altered association with the dopamine transporter, compared with WT σ₁R (Hong et al., 2017), and exerted dominant-negative effects in disrupting association of WT σ₁R with μ opioid receptors (Pan et al., 2017). These results suggest that dysregulation of σ₁R quaternary structure impairs its physiologic function.

Intriguingly, buffer pH significantly altered the oligomeric states of σ₁R (Fig. 7E). Although the underlying mechanism and physiologic significance are beyond the scope of this study, a rudimentary inference is that σ₁R multimers dissociate in acidic lysosomes to facilitate degradation. This serendipitous finding helps to examine how σ₁R oligomerization affected ligand binding. Results (Fig. 7) indicate that σ₁R multimers possess high-affinity and high-capacity binding of [³H](+)-pentazocine. Because Δ36aa σ₁R formed monomer exclusively but lacked binding (Table 1), it was inferred that monomeric σ₁R could not bind (+)-pentazocine. This hypothesis is consistent with a previous study using σ₁R mutants from bacterial expression (Gromek et al., 2014).

In summary, this study demonstrated that multiple domains coordinate the oligomerization of σ₁R. The equilibrium between monomers, dimers, and multimers of σ₁R is dynamically regulated by agonists or antagonists in distinct manners. Antagonists promote σ₁R multimerization, whereas agonists facilitate its dissociation (Fig. 8). σ₁R mutants implicated in neurodegenerative diseases displayed aberrant multimerization, suggesting that balance in σ₁R oligomerization is important in its physiologic function. Extensive studies have shown that σ₁R associates with a plethora of partner proteins that are involved in diverse cellular signaling pathways. Distinct regulation of σ₁R multimerization by agonists and antagonists may selectively modulate activities of clientele proteins within their interactome network. Future work will shed light on whether σ₁R oligomerization may be precisely controlled by cells’ adaptive responses to physiochemical changes in the environment, which in turn may impact these signaling mechanisms.