2. Intracellular Localization of the S1R
S1R primarily resides in the ER membrane where it forms microdomains
[42,59,61,62][42][59][60][61]. Its localization is in contrast to the uniform distribution pattern of ER markers, such as the Sec61b protein. A significant proportion of S1R is localized to MAMs, an ER sub-compartment closely associated with the mitochondria
[42[42][59],
59], in proximity to lipid droplets
[63][62], and at the ER-plasma membrane (PM) junctions
[59,64][59][63]. It is likely that S1R are localized to additional inter-organelle contact sites, but this has not been systematically investigated.
MAMs are distinct from the rest of the ER as they contain enzymes involved in lipid synthesis, calcium signaling, cholesterol metabolism, and the ER stress-response pathways
[65,66,67,68,69][64][65][66][67][68]. A detailed protein composition of MAMs was initially characterized by biochemical purifications
[70,71][69][70] and more recently established using sophisticated proximity labeling approaches
[72,73,74][71][72][73].
While the precise lipid composition of the MAMs has not yet been elucidated, recent evidence suggests that cholesterol and ceramide content is significantly higher in MAMs compared to the rest of the ER
[57,75,76][57][74][75]. Therefore, MAMs can be thought of as specialized ER signaling domains characterized by unique protein and lipid compositions, similar to PM lipid rafts
[60][76].
While PM lipid and protein heterogeneity was visualized using the giant plasma membrane-derived vesicle technique
[77,78,79][77][78][79], only recently was a similar method to yield endomembrane-derived giant unilamellar vesicle (GUV)-like vesicles developed
[80]. Using this approach, it was observed that certain, but not all ER contact sites (such as ER-mitochondria, ER-PM, and ER-lipid droplets) showed separation of the glycosylphosphatidylinositol (GPI) ER-targeted marker with a strong affinity for lipid-ordered phase
[80]. Similar lipid and protein compartmentalization were recently observed at the inter-organelle contact sites in yeast
[81], providing additional experimental evidence for microscopic lipid heterogeneity in the ER.
Binding immunoglobulin protein/glucose-regulated protein 78 (BiP/GRP78) was identified as a major S1R binding partner using pull-down experiments
[42]. The S1R interacts with BiP in a calcium- and agonist-dependent manner. At high Ca
2+ concentrations or in the absence of an agonist, the S1R forms a complex with BiP, keeping it in an inactive state. Therefore, BiP interaction contributes to S1R retention in the ER.
Under conditions of calcium depletion, or agonist activation, the S1R dissociates from BiP
[42]. Agonist stimulation leads to redistribution of S1R from clusters to the ER, plasma membrane, and extracellular space
[55,61][55][60]. In flotation assays, activation by an agonist causes the S1R to translocate from detergent-resistant to detergent-soluble fractions
[55,61,63][55][60][62].
S1R activation releases the inhibitory interaction with BiP and allows for the S1R to interact with various partners inside and outside of the MAMs, including the inositol-1,4,5-triphosphate receptor type 3 (InsP
3R3)
[34,42][34][42]. Our recent results
[59] and previous studies
[57] suggest that a direct, high-affinity association of the S1R with cholesterol and ceramides may also contribute to S1R targeting by MAMs.
3. Interaction of S1R with ER Membranes
S1R was shown to interact with cholesterol in vitro
[55], suggesting that S1R association with cholesterol plays an important role in MAM targeting of the S1R
[57] and in modulation of PM cholesterol levels
[58].
Recently, we demonstrated that cholesterol promotes the formation of S1R domains in a model lipid bilayer system
[59]. Using GUVs with reconstituted fluorescent-labeled S1R, we observed that cholesterol was sufficient to cause clustering of recombinant S1R in the absence of any other proteins
[59]. This study further identified a novel cholesterol-binding site within the transmembrane (TM) domain of the S1R. Additionally, point mutations in the TM domain, which weaken the interaction of the S1R with cholesterol, result in the impaired redistribution of S1R into the entire ER network
[59].
In vitro, S1R clustering was observed in a narrow range between 2.5% and 5.0% mol cholesterol
[59], comparable with the cholesterol-dependence of sterol regulatory element binding protein-2: sterol regulatory element-binding protein cleavage-activating protein complex (SREBP-2:Scap)
[82]. Our data suggests that S1R actively participates in the assembly of micrometer-size cholesterol-enriched microdomains. On the basis of these observations, we propose that S1Rs promote the formation and stabilization of MAM microdomains, and potentially other ER contact sites
[59]. Consistent with this idea, the genetic deletion of S1R impairs MAM stability and results in a reduced number of contacts, as observed by electron microscopy (EM) and biochemical fractionation
[43].
The S1R has an unusually long, single transmembrane domain
[59]. Local “measurements” of bilayer thickness with transmembrane sensors showed that bilayers surrounding S1R domains are thicker
[59]. In the crystal structure, the S1R molecules are organized as trimers, with their C-terminal ligand-binding domain partially embedded in the membrane
[31,83][31][83]. These amphipathic helices at the C-terminus of the S1R are rich in aromatic residues and, thus, can play a role in additional stabilization of the bilayer structure, a phenomenon previously observed for other amphipathic helices
[84,85][84][85].
Thus, we reasoned that the local membrane thickness is increased in S1R-formed ER microdomains
[59], which likely plays a role in the sorting of ER membrane proteins to these domains
[79,81][79][81].
4. The S1R as a Therapeutic Target for the Treatment of Neurodegenerative Diseases
The S1R is a well-established target for the treatment of neurodegenerative disorders, and it plays a key role in neurodegenerative diseases. Several S1R mutations have been identified to be associated with ALS and frontotemporal dementia (FTD)
[97][86]. Two complete loss of function (LOF) mutations cause a juvenile form of ALS
[12,43][12][43]. However, missense mutations that partially impair protein function are associated with an adult form of ALS, showing a dose response between the function of S1R and the age of onset of ALS
[98][87]. Additional LOF mutations in S1R cause hereditary motor neuropathies
[15,18][15][18]. Furthermore, some variants of the S1R gene are associated with increased risk for Alzheimer’s disease (AD)
[99][88]. In addition, S1R expression levels are reduced in sporadic ALS
[20], Parkinson’s disease (PD), and Alzheimer’s disease (AD) patients
[21,100][21][89].
Further support for the role of S1R in neurodegenerative diseases comes from animal models. In preclinical models, genetic ablation of S1R (S1R−/−) in mice exacerbates pathology and phenotypic presentation of several neurological disorders. For example, S1R−/− mice display impairments in motor function and degeneration of motor neurons at 5 months of age
[101][90]. AD mice (APP
sweInd), which lack S1R, show enhanced behavioral and cognitive impairments, as well as a significant reduction in the levels of the brain-derived neurotrophic factor (BDNF) compared to APP
sweInd mice expressing the S1R
[102][91]. In the ALS SOD1
G93A mouse model that also lacks S1R expression (SOD1
G93A/S1R KO), disease progression is accelerated, as revealed by earlier weight loss and by a ~32% decrease in survival time relative to SOD1
G93A mice with normal S1R expression
[101][90]. These results suggest that the S1R plays an important role in healthy neuronal physiology.
S1R activation by agonists has demonstrated neuroprotective effects in multiple cellular and animal models of neurodegeneration (reviewed in
[5,45,103,104][5][45][92][93]). Extensive evidence suggests that the mechanisms responsible for the neuroprotective effects of S1R agonists include the stabilization of Ca
2+ signaling
[42,44[42][44][94][95],
105,106], an increase in the secretion of BDNF and the potentiation of BDNF-tropomyosin-related kinase B (TrkB) signaling
[107,108[96][97][98][99][100][101],
109,110,111,112], the stimulation of cyclic AMP-responsive element-binding protein (CREB)-mediated transcription
[110][99], changes in the activity of the plasma membrane ion channels and receptors
[113[102][103][104][105],
114,115,116], the potentiation of the N-Methyl-D-aspartate (NMDA) receptor response
[113[102][106][107][108],
117,118,119], and an improvement in mitochondrial function
[43,120,121,122][43][109][110][111].
For example, we and others have shown that the highly selective and potent S1R agonist pridopidine restores the dysregulated ER Ca
2+ signaling and enhances spine density in Huntington’s disease (HD) and Alzheimer’s disease (AD) neurons
[44,45,123][44][45][112]. Furthermore, S1R activation by pridopidine enhances synaptic plasticity in HD cortical neurons
[123][112] and exhibits a robust neuroprotective effect against mutant huntingtin-induced cell death in mice’s primary HD neurons and in HD patient-derived induced-pluripotent stem cells (iPSCs)
[124][113]. Pridopidine has been found to upregulate BDNF secretion, potentiate BDNF-TrkB signaling, and enhance BDNF axonal transport in several different models of neurodegenerative diseases including HD and ALS
[109,110,125][98][99][114]. Pridopidine has shown protective effects on several mitochondrial functions in various human and mouse models of HD. In primary HD neurons, pridopidine enhances mitochondria-ER tethering and restores mitochondrial function as measured by increased ATP production, respiration, and mitochondrial membrane potential
[126][115]. All these effects are exquisitely mediated by the activation of the S1R, as either a genetic deletion of the S1R or a pharmacological inhibition using an S1R antagonist, completely abolishes pridopidine’s neuroprotective effects, as shown in the studies mentioned above
[126][115].
Recent clinical studies have shown the potential efficacy of the selective S1R agonist pridopidine in HD patients, demonstrating maintenance or slowing the decline of the patient’s functional capacity
[127,128][116][117]. The non-selective S1R/Muscarinic (M1R) agonist blarcamesine shows a potential beneficial effect in AD
[129][118]. Clinical pivotal studies with pridopidine are currently ongoing for HD and ALS (NCT04556656, NCT04297683). Blarcamesine is currently being evaluated for AD, Rett syndrome, and PD dementia patients (NCT04314934, NCT04304482, NCT04575259). Results of completed clinical trials of S1R agonists in variety of disorders have been comprehensively summarized in recent reviews
[4,130][4][119].
How can the activation of S1R exert such pleotropic and variable effects on cellular signaling? We propose that the agonist activation of S1R results in the remodeling of S1R microdomains (). In our experiments
[59] and in published studies
[55[55][60],
61], the activation of S1R has resulted in the disassembly of the S1R oligomers. We propose that agonists cause partial disassembly and remodeling of S1R microdomains in the ER, leading to a rapid release of mature proteins that are trapped in these microdomains.
Consistent with this hypothesis, increases in the levels of PM proteins are often observed following S1R stimulation with an agonist. For example, the S1R agonist SKF-10047 increased the plasma membrane fraction of the GluN1, GluN2A, and GluN2b NMDAR subunits
[118][107]. Cocaine increased the PM fraction of Kv1.2
[50,51][50][51]. Cell surface expression of programmed cell death 1 ligand 1 (PD-L1) was increased after the agonist stimulation, and lower levels of PD-L1 were observed in S1R knockdown (KD) cells
[131][120]. On the other hand, S1R knockdown (KD) decreased the stability and levels of the mature human ether-à-go-go-related gene (hERG), as well as hERG currents
[132][121]. A shorter protein half-life was observed for InsP
3R3 in MAMs in S1R KD cells
[42], and the turnover of p35, which is dependent on protein myrostyilation, was lower in S1R deleted cells
[133][122]. In contrast, however, the stability of UDP-galactose:ceramide galactosyltransferase (UGT8) was increased in S1R deleted cells
[134][123].
The same idea applies not only to PM proteins, but also to secreted proteins. BDNF release increased after treatment with the S1R agonists pridopidine and cutamesine in B104 cells and in astrocyte cultures
[111,135][100][124]. Levels of secreted BDNF and glial cell-derived neurotrophic factor (GDNF) were increased in substantia nigra after pridopidine and sigma receptor agonist PRE-084 treatment in the experimental Parkinsonism model
[112,136][101][125]. Moreover, it was shown that BDNF processing is modulated by S1R at the post-translational stage
[135][124], in line with our hypothesis.
On the basis of these results, we propose that there is a “reserve pool” of PM and secreted proteins which accumulate in cholesterol-rich ER microdomains. S1R agonists facilitate the remodeling of these microdomains and the rapid release of PM and secreted proteins, resulting in a robust response to stress and in neuroprotective effects.
Additional neuroprotective effects of S1R activation may also be related to the modulation of ER Ca
2+ signaling, in particular inositol trisphosphate receptor (InsP
3R)-mediated Ca
2+ signaling
[42,44,137][42][44][126] and ER-mitochondrial Ca
2+ transfer
[42,43][42][43]. The possible explanations for the observed potentiation of InsP
3-induced Ca
2+ release
[137,138,139][126][127][128] can include the direct effects of cholesterol on the activity of the InsP
3Rs, the modulation of ER Ca
2+ content through cholesterol regulation of sacro/endoplasmic reticulum Ca
2+ (SERCA) pump
[140[129][130][131],
141,142], the removal of the ankyrin inhibition of InsP
3Rs
[137][126], and the redistribution of InsP
3Rs from MAMs to the peripheral ER
[42,43][42][43].
This proposed model may explain the pleiotropic effects of S1R agonists and provide appropriate context for the development of S1R-targeting therapeutic agents.