Activation mechanisms and structural dynamics of STIM proteins

The family of stromal interaction molecules (STIM) includes two widely expressed single‐pass endoplasmic reticulum (ER) transmembrane proteins and additional splice variants that act as precise ER‐luminal Ca2+ sensors. STIM proteins mainly function as one of the two essential components of the so‐called Ca2+ release‐activated Ca2+ (CRAC) channel. The second CRAC channel component is constituted by pore‐forming Orai proteins in the plasma membrane. STIM and Orai physically interact with each other to enable CRAC channel opening, which is a critical prerequisite for various downstream signalling pathways such as gene transcription or proliferation. Their activation commonly requires the emptying of the intracellular ER Ca2+ store. Using their Ca2+ sensing capabilities, STIM proteins confer this Ca2+ content‐dependent signal to Orai, thereby linking Ca2+ store depletion to CRAC channel opening. Here we review the conformational dynamics occurring along the entire STIM protein upon store depletion, involving the transition from the quiescent, compactly folded structure into an active, extended state, modulation by a variety of accessory components in the cell as well as the impairment of individual steps of the STIM activation cascade associated with disease.


Introduction
A variety of ions such as calcium, potassium, magnesium, chloride and sodium are essential for all known forms of biological life.Among them, the role of Ca 2+ is of special importance.Ca 2+ signalling is exploited by complex organisms to maintain healthy body functions such as the formation of bones and teeth, the control of immune response and muscle contraction.In addition, single cells depend on Ca 2+ to function as a second messenger to govern processes including cellular growth and differentiation, secretion, migration, gene expression, fertilization and apoptosis (Berridge, 1995(Berridge, , 2000(Berridge, , 2003;;Brini et al., 2013b;Parekh, 2010;Putney, 1986Putney, , 2005)).Due to the importance of these processes for cellular homeostasis, tight regulation of spatial and temporal Ca 2+ distribution is essential.Cells achieve this through a complex interplay of Ca 2+ -sensing, Ca 2+ -buffering and Ca 2+ -transporting proteins.In resting cells, these molecular machines ensure the maintenance of high Ca 2+ in the organelles and on the extracellular side, signalling and combines live-cell experiments with computer simulation.In addition, his lab focuses on bioelectronic medicine, including optoelectronic neuro-stimulation and electronically controlled local chemotherapy.Isabella Derler is an assistant professor at the Biophysics Institute at the JKU Linz.Her laboratory explores the molecular and cellular mechanisms by which store-operated Ca 2+ channels are activated as well as cellular factors that modulate their function.
whereas a low Ca 2+ concentration is established in the cytosol.Upon cellular signalling events, this Ca 2+ toolkit is required to create different patterns of Ca 2+ concentration enhancements including sustained, transient or oscillatory forms, ensuring Ca 2+ ion signalling versatility (Brini et al., 2013a(Brini et al., , 2013b;;Parekh & Putney, 2005).Cytosolic Ca 2+ concentrations can be increased either by the depletion of the intracellular Ca 2+ stores or the influx of Ca 2+ from the extracellular space (Putney, 2005).One Ca 2+ entry route which connects these two events is called store-operated Ca 2+ entry (SOCE).
A classical type of store-operated Ca 2+ ion channel is the Ca 2+ release-activated Ca 2+ (CRAC) channel, formed by a Ca 2+ sensor protein, stromal interaction molecule (STIM), in the ER membrane and the Ca 2+ ion channel, Orai, in the PM (Bonhenry et al., 2019;Fahrner et al., 2009Fahrner et al., , 2017;;Feske et al., 2006;Hoth & Penner, 1992;Liou et al., 2005;Prakriya & Lewis, 2015;Roos et al., 2005;Schober et al., 2019b;Vig et al., 2006;Zhang et al., 2006).Orai consists of four transmembrane (TM) domains with both the N-and C-termini located in the cytosol.A total of six Orai monomers are required to form a functional channel pore.The TM domains thereby assemble as concentric rings around the ion-conducting pore in the centre, whereby the innermost ring that actually defines the pore is established by the six TM1 domains (Cai et al., 2016;Hou et al., 2012Hou et al., , 2018Hou et al., , 2020;;Liu et al., 2019;Yen et al., 2016).The single TM domain protein STIM contains an EF-hand as part of its ER luminal N-terminus that forms the basis of its Ca 2+ sensing ability by allowing it to bind Ca 2+ in a concentration-dependent manner (Fahrner et al., 2020;Stathopulos & Ikura, 2010).When the ER Ca 2+ concentration decreases as a result of ER store depletion, Ca 2+ dissociates from the STIM EF-hand and thereby triggers unfolding and dimerization of the STIM N-terminus (Stathopulos et al., 2006).The dimerization signal propagates via zipping of the TM domain and the cytosolic C-terminal domains, where a series of conformational changes occur that transform the whole C-terminus from a quiescent, tight state into an active, extended state (Fahrner et al., 2014;Fahrner, Stadlbauer et al., 2018;Hoglinger et al., 2021;Ma et al., 2015;Muik et al., 2009Muik et al., , 2011;;Rathner et al., 2021;Subedi et al., 2018;van Dorp et al., 2021;Zhou et al., 2013).Homo-oligomerization and cluster formation accompany this process.The STIM clusters, which are also referred to as puncta, localize to ER-PM junctions.Here, the two membranes are sufficiently close for the extended STIM to bridge the distance between them and physically bind to and activate Orai via the latter's C-and N-termini as well as loop2, completing the SOCE activation process (Fig. 1) (Bakowski et al., 2021;Barak & Parekh, 2020;Derler et al., 2013;Fahrner et al., 2017;Fahrner, Pandey et al., 2018;Frischauf et al., 2016;Lewis, 2020;Luik et al., 2008;Muik et al., 2008;Park et al., 2009;Yeh & Parekh, 2018;Yuan et al., 2009;Zhou et al., 2010).Besides the common route of store-operated activation, STIM1 can also be activated in a store-independent manner by other factors such as oxidative stress or temperature change, which are reviewed in detail elsewhere (Mancarella et al., 2011;Mancarella et al., 2013;Prakriya & Lewis, 2015;Soboloff et al., 2011).
In this review, we discuss remarkable conformational changes in STIM proteins and highlight the dynamics of structural changes during its activation.We begin with a description of the protein family, including the general structure as well as the similarities and differences between the individual STIM variants.The focus will subsequently shift to in-depth characterizations of STIM proteins in the quiescent and active states that are assumed in the presence of high and low ER Ca 2+ levels, respectively.We conclude with a discussion of diseases associated with STIM variants as well as an overview of STIM modulatory proteins and other regulatory factors.The family of STIM proteins STIM proteins are dimeric, type I single-pass TM proteins anchored mainly in the ER membrane, but they are also found in acidic stores or to a minor extent in the PM (Jardin et al., 2013;Liou et al., 2005;Lunz et al., 2019;Roos et al., 2005;Sabbioni et al., 1999;Spassova et al., 2006;Zbidi et al., 2011;Zhang et al., 2005).They consist of an N-terminal portion within the ER lumen and a considerably larger C-terminal portion within the cytosol.Aside from their function as highly specialized sensors of ER luminal Ca 2+ , STIM proteins are also equipped with structural domains to homomerize as well as to bind to and gate Ca 2+ -selective Orai channels in the PM (Fahrner et al., 2017;Feske et al., 2006;Liou et al., 2005;Park et al., 2009;Prakriya & Lewis, 2015;Prakriya et al., 2006;Roos et al., 2005;Yuan et al., 2009).
Humans possess two homologous, structurally similar STIM proteins called STIM1 and STIM2, which play an indispensable role in the immune system (Shaw & Feske, 2012;Wissenbach et al., 2007) and are ubiquitously expressed in a variety of other tissues (Williams et al., 2001).The main functional difference is a reduced Ca 2+ affinity in STIM2, which increases its sensitivity to small drops of the ER Ca 2+ concentration.For this reason, STIM2 already localizes to ER-PM junctions when the Ca 2+ stores deviate slightly from the resting level and fine-tunes the rate of Ca 2+ entry by triggering a minor but sustained activation of SOCE.In contrast, STIM1 is only activated upon major Ca 2+ store depletion leading to strong but transient activation of SOCE, swiftly returning the Ca 2+ store to resting levels (Brandman et al., 2007;Liou et al., 2007;Zheng et al., 2008Zheng et al., , 2011;;Zheng et al., 2018;Zhou et al., 2009).
STIM1L is the best studied STIM1 isoform.Discovered in 2011, it was found to harbour an additional 106 amino acid-long actin-binding domain (ABD) downstream of ID-STIM due to inclusion of exon 13 (termed 'L').The ABD was shown to allow STIM1L to permanently localize to ER-PM junctions for interaction with Orai1 channels.STIM1L expression was detected in neonatal rat cardiomyocytes, brain and heart of mice as well as human skeletal muscle cells.Due to this, STIM1L is believed to be required for rapid SOCE activation in the range of seconds that can be observed in skeletal muscle cells (Darbellay et al., 2011;Horinouchi et al., 2012;Luo et al., 2012), in contrast to other cells where SOCE is an inherently slow process with activation times in the range of a minute (Parekh & Putney, 2005).This notion is supported by a recent study showing that knockdown of STIM1L has negative effects on myoblast differentiation and leads to a decreased size of myotubes formed by them (Antigny et al., 2017).
Very recently, two other STIM1 isoforms named STIM1A and STIM1B were reported (Knapp et al., 2022;Ramesh et al., 2021).STIM1A arises from inclusion of exon 11 (termed ' A') into the STIM1 sequence.This means that a 31 amino acid-long domain A is inserted immediately downstream of ID-STIM.STIM1A is mainly expressed in heart, kidney, testis and astrocytes.It was found to reduce SOCE activation although no differences from wild-type STIM1 were detected in terms of clustering and binding to Orai1 channels (Knapp et al., 2022).STIM1B, in contrast, is produced by inclusion of another exon 13 variant (termed 'B') into the STIM1 sequence.Although a new 26 amino acid-long domain B is therefore inserted downstream of ID-STIM, STIM1B is considerably shorter than conventional STIM1 due to a frameshift after domain B leading to premature termination of translation.STIM1B has so far been exclusively detected in neuronal tissues and was demonstrated to attenuate SOCE activation and reduce slow Ca 2+ -dependent inactivation (SCDI).Domain B targets STIM1B to presynaptic ER, where it promotes synaptic signalling through increased replenishment of synaptic vesicles (Ramesh et al., 2021).
The STIM2.1 isoform contains an 8 amino acid-long insert (VAASYLIQ) in the CAD/SOAR CC2 domain due to inclusion of an additional exon 9 into the STIM2 sequence.STIM2.1 was shown to act as a SOCE inhibitor (Miederer et al., 2015;Rana et al., 2015).Information about the STIM2.3isoform is very limited.It is generated by inclusion of an alternative exon 13.This triggers a premature termination of translation and results in a protein that is 148 amino acids shorter than conventional STIM2.The expression of STIM2.3 seems to be limited and its function remains elusive to date (Miederer et al., 2015).The limited expression of STIM2.3 might be an indicator that, much like STIM1B, it is only required by certain tissues to fulfill specialized tasks that cannot be fulfilled by other STIM variants.Due to the missing EB domain and PBD of STIM2.3, its intracellular trafficking as well as clustering probably differ from the other STIM2 isoforms.It is conceivable that STIM2.3 is a better SOCE activator than conventional STIM2, as microtubule binding was shown to attenuate the ability of STIM proteins to migrate to ER-PM junctions for interaction with Orai channels (Chang et al., 2018;Wang et al., 2018).
All in all, the STIM isoforms considerably expand and diversify the capabilities of the protein family.Different cell types harness alternative splicing to target STIM proteins to specific intracellular contact sites, where they subsequently shape and modulate SOCE according to their individual needs.

Key features of STIM in the resting state
STIM N-terminus under resting conditions.As described in the introduction part, the functionally relevant N-terminal part of STIM contains three highly conserved structural segments: a canonical and a non-canonical EF-hand motif as well as a SAM domain.They act as a Ca 2+ sensor, reacting to changes in Ca 2+ concentration in a range between 100 and 400 μM (Brandman et al., 2007;Luik et al., 2008) (half-maximum activation of STIM1 is 200 μM and of STIM2 is 400 μM).Nuclear magnetic resonance (NMR) spectroscopy structures of Ca 2+ -bound STIM1 and STIM2 N-terminal portions resolved paired EF-hand structures forming a hydrophobic pocket embracing and stabilized by the SAM domain (Stathopulos et al., 2008;Zheng et al., 2011), resulting in a compact globular structure (Fig. 2).In detail, Ca 2+ -loaded recombinantly expressed EF-SAM STIM1 proteins showed monomeric EF-SAM existence (17.4 kDa) and relatively low Ca 2+ binding affinity with an apparent K d of 0.2-0.6 mM.At the structural level, the canonical EF-hand forms a helix-loop-helix EF-hand motif, to which a Ca 2+ ion is bound (Stathopulos et al., 2008).The second non-canonical EF-hand helix-loop-helix stabilizes the canonical EF-hand between loop regions of each motif, forming an antiparallel β-sheet structure.The two EF-hands are bound to the SAM domain, which folds into a five-helix bundle, via a connecting short linker helix.The hydrophobic pocket of the compactly folded EF-SAM complex buries 11 residues (V68, I71, H72, L74, M75, L92, L96, K104, F108, I115 and L120) of the EF-hand and is stabilized by L195 and L199 of the SAM domain of STIM1 (Fig. 2A) (Stathopulos et al., 2006(Stathopulos et al., , 2008)).Negatively charged residues of the canonical EF hand, non-canonical EF hand and SAM domain (canonical EF: D77, D78, D82, D84, E86, E87, D89, E90, E94, D95; non-canonical EF: E111, D112, E118, E128; SAM: D135, E136, D196; Stathopulos et al., 2008) facing the surface of this compactly folded globular domain likely prevent intermolecular interactions with other N-terminal STIM1 domains.Despite the slightly higher Ca 2+ affinity of STIM1 in comparison to STIM2, NMR spectroscopy revealed similar structural features of STIM2 (Brandman et al., 2007;Zheng et al., 2011).Zheng et al. (2008;2011) reported a highly α-helical, compact EF-hand-SAM structure of Ca 2+ -loaded STIM2, similar to STIM1.Like STIM1, hydrophobic residues of STIM2 (L72, I75, H76, M79, F85, M100, L103, L108, L112, I119, L142 and W128) form a non-polar cleft, which docks the hydrophobic C-terminal part of the SAM domain (Fig. 2B).The most drastic difference between the STIM1 and STIM2 structure was observed for the positioning of the SAM domain with respect to other N-terminal regions.The STIM2 structure revealed a parallel conformation of the canonical EF-hand α2 and SAM domain α10 helix, whereas a perpendicular configuration was observed for these two helices in STIM1 (Fig. 2A and B).This allows for additional hydrophobic and interdomain ionic interactions of the STIM2 EF-SAM structure, promoting its stability together with differences in the composition of the SAM domain cores.The latter buries 12 non-polar residues in STIM2 compared to nine hydrophobic residues for STIM1 (Fig. 2A and B).These structural differences were further proven to be functionally relevant by STIM1-STIM2 chimeras, where constructs including the STIM2 SAM domain showed the highest stability and most resistance to Ca 2+ depletion-induced oligomerization.The lower Ca 2+ affinity and higher stability of the EF-SAM domain of STIM2 seems to be an important determinant for the distinct function of STIM2, indicating that already fractional changes in the STIM2 architecture could have a drastic influence on the activation state of the protein (Zheng et al., 2011).
In addition to the structures for human STIM1 and STIM2, Enomoto et al. (2020) determined the X-ray crystal structure of the C. elegans STIM luminal EF-SAM domain.Upon Ca 2+ saturation, the canonical helix-loop-helix motif coordinates one single Ca 2+ ion in a pentagonal bipyramidal geometry which is followed by a second non-canonical helix-loop-helix structure, a short linker helix and again a five-helix-bundle SAM domain.Like for STIM1 and STIM2, C. elegans STIM also forms a hydrophobic cleft with a docking site for hydrophobic residues of SAM α10 (Fig. 2C).Notably, an additional ionic bond between R85 of the non-canonical EF-hand and D177 of the SAM domain further stabilizes the compactly folded, Ca 2+ -bound state.The α2 and α10 orientation of C. elegans is similar to that of STIM2 with a nearly parallel orientation.A remarkable feature of the structure of C. elegans was the discovery of a non-conserved α-helical segment upstream of the EF-SAM core, which is stabilized by an intrahelix ionic interaction between E36 and R39 (Fig. 2C).All these additional contact sites of the EF-hand-SAM domain together account for the highest stability of any STIM protein studied to date (Enomoto et al., 2020).How and whether the analogous, non-conserved N-terminal region of human STIM proteins is involved in stabilization processes remains to be clarified.
Binding of Ca 2+ to the N-terminal part of STIM proteins is important to keep the protein inactive, yet the exact number of bound Ca 2+ ions upon rest is still under debate.Structural analysis studies reported that a single Ca 2+ ion is coordinated by the first helix-loop-helix region of the canonical EF-hand (Stathopulos et al., 2006(Stathopulos et al., , 2008;;Zheng et al., 2011).Gudlur et al. (2018) stated, based on isothermal titration calorimetry and fluorescence competition assay for Ca 2+ binding, that five to six sites per monomer are occupied by Ca 2+ ions during resting state conditions.They also examined the number of Ca 2+ ions bound using an 11-residue mutation cluster introduced in the canonical EF-hand domain of STIM1, which inhibited the ability to form inducible puncta.Introduction of only a subset of these residues (E94Q/D100N/E111Q/D112 N/E118Q/D119N) into an EF-SAM Thermus thermophilus Gro-P-like protein E (GrpE) fusion construct showed both constitutive as well as store operated STIM1 activation, which they attributed to weaker Ca 2+ binding affinity leading to five to six bound Ca 2+ ions.A molecular dynamics simulation approach by Schober et al. (2019a) additionally suggested that Ca 2+ does not only bind to the canonical EF-hand but associates also with the non-canonical EF-hand and SAM domain.However, we have stated that a single bound Ca 2+ ion, which remains captured throughout the whole simulation by the side chains of D76 and D78 of STIM1, is sufficient to keep the protein in a stable, compactly folded resting conformation.Indeed, this is in line with previous studies showing that a D76A mutation within STIM1 impacts the Ca 2+ sensitivity of the EF-hand domain, switching the protein into a constitutively active state (Liou et al., 2005;Spassova et al., 2006;Zhang et al., 2005).Analogous studies on STIM2 might potentially help to find a reason for its lower Ca 2+ affinity compared to STIM1.
STIM1 TM domain under resting conditions.Following the STIM N-terminal part, within the STIM1 TM domain, individual amino acids as well as intra-dimeric interactions and conformations contribute to the control of the resting state (Hirve et al., 2018).Using a tryptophan scanning approach (Ma et al., 2015), two amino acids (I220, C227) were identified within the TM segment that maintain the inactive state of STIM1.Consistently, their mutation (I220W, C227W) led to constitutive STIM1 activity.Biochemical, spectroscopic and computational approaches have shown that mainly the C-terminal segments (aa 221−232) of the two TM domains interact, while their N-terminal segments (aa 212−220) are separated by a crossing angle of about 45°in the resting state (Ma et al., 2015).However, recent single molecule Förster resonance energy transfer (FRET) analysis of the cytosolic part (van Dorp et al., 2021), as described in more detail below, suggests a clear separation of the TM domains of a STIM1 dimer in the resting state, and hence previous results may need to be reviewed with the novel insights of the FRET approach.
In the quiescent state, STIM1 proteins adopt a dimeric configuration which is maintained by intra-and intermolecular interactions of their coiled-coil regions (Covington et al., 2010).Several key sites critical for dimer formation were characterized by the crystal structure of human SOAR and the NMR structure of CC1α3-CC2.The crystal structure of human SOAR consists of two monomers that are antiparallel to each other and adopt a V-shaped conformation.Each monomer exhibits an R-like structure formed by the four α-helices Sα1-Sα4.In this structure, hydrogen and hydrophobic bonding, as well as stacking interactions between the N-terminal residues of one SOAR (L347, W350, L351, T354, Y361 in α1) and the C-terminal residues of another (R429, W430, I433, L436 in α4) apparently contribute to STIM1 dimer formation (Fig. 3A).Furthermore, mutagenesis studies confirmed their importance as their mutation negatively impacts oligomerization and SOAR-Orai1 coupling (Yang et al., 2012).The NMR structure of two overlapping SOAR domains revealed interactions of the antiparallel oriented monomers comprising CC1α3 (aa 320−331) and CC2 (aa 355−369), specifically of V324-A327 and E358-K366 (Stathopulos et al., 2013).At the time of publication of these structures, it was still unclear whether they represent a quiescent, intermediate or active state.A recent single-molecule FRET study unveiled several contact sites which are more similar to the SOAR crystal structure than to the CC1α3-CC2 NMR structure, indicating that the contact sites observed in the crystal structure in particular occur in the quiescent state (van Dorp et al., 2021).However, to what extent the CC1α3-CC2 NMR structure fits into the STIM1 activation cascade is currently unclear.It may represent an intermediate state during the activation process.Thus, further structural studies of STIM1 fragments or full-length STIM1 are highly awaited.
A main interaction site maintaining the resting conformation of the STIM1 C-terminus represents the inhibitory clamp formed by its coiled-coil region.Indeed, in initial attempts, STIM1 activation was achieved by neutralizing acidic residues in CC1α3 leading to the conclusion that these residues bind to basic residues in CAD, thus forming an inhibitory clamp (Korzeniowski et al., 2010).In accordance, the crystal structure of the coiled-coil region of ceSTIM1 C-terminus unveiled an inhibitory helix interacting with ceSOAR.Functional characterization of the analogue region in human STIM1, located at the end of CC1α3, revealed that this region maintains the quiescent state of STIM1 (Yang et al., 2012).However, because the CC1α3-CAD fragment remained active (Zhou et al., 2013) and a STIM1 deletion mutant lacking CC1α3 resulted in weak Orai1 activation under J Physiol 602.8 resting conditions (Fahrner et al., 2014), it was concluded that CC1α3 alone was not responsible for sustaining the inactive state.Later studies uncovered that CC1α1 and CC3 form the inhibitory clamp which retains STIM1 in the quiescent state (Fahrner et al., 2014;Fahrner, Stadlbauer et al., 2018;Muik et al., 2011;Shrestha et al., 2022;Zhou et al., 2013).Supportively, a double-labelled OASF fragment showed robust FRET (Muik et al., 2011).A switch to low FRET upon incorporation of single point mutations in CC1α1 (e.g.L248S, L251S, L258S)  2021).The whole STIM1 protein was reconstructed using Modeller software (Fiser & Sali, 2003), which allowed integration of the different subdomains and connecting of the luminal and cytosolic domains with a modelled TM domain.The TM domain and CC1α1 were modelled as a continuous α-helix, up to residue 271.This new subdomain was then apposed to CC3 from SOAR to maximize the contact surface between hydrophobic residues.CC1α2 and CC1α3 were also modelled as α-helices and positioned on top of the CC1α1/SOAR construct.The short sequences of amino acids in between were considered unstructured and modelled as random coils.The CC1α3 domain was then connected to the SOAR crystal structure by adding residues 337-345 in a configuration corresponding to an α-helix.To generate a dimer, SOAR of the first STIM1 monomer was used to position SOAR of the second STIM1 monomer.Specifically, the dimer was created by aligning the SOAR Cα atoms from the first fully reconstructed monomer to the SOAR Cα atoms of the second monomer.A short MD simulation was then performed where all heavy atoms from the backbone remained fixed to relax the side chains.Subsequently, a 1 ns-long molecular dynamics simulation without any constraints was performed to relax the structure.G, graphical representation of residues identified as being important for STIM dimerization (from TM domain in violet to CC1α1, CC1α2 and CC1α3 in blue).
or CC3 (e.g.L416S, L423S) indicated the release of the tight conformation of the STIM1 fragment (Muik et al., 2011;Zhou et al., 2013).Moreover, a two-component FRET pair composed of two distinctly labelled STIM1 fragments, namely (i) STIM1 truncated after CC1 and (ii) SOAR, exhibited coupling only under resting cell conditions (Shrestha et al., 2022).Consistently, a method which detects FRET-derived interaction in a restricted environment (FIRE) revealed a clear interaction of CC1α1 and CC3 (Fig. 3B), which could be abolished by the above-mentioned mutations (Fahrner et al., 2014).In contrast to STIM1, OASF of STIM2 exhibits a more expanded, open conformation likely responsible for clustering of STIM2 with minimal store depletion.This STIM2 C-terminal conformational peculiarity assists STIM1 to transition to the active state, allowing STIM1/Orai1 activation (Subedi et al., 2018).However, the structural properties of STIM2 OASF are still elusive.Notably, sequence comparison revealed only minor differences between STIM1 and STIM2 (Dziadek & Johnstone, 2007;Grabmayr et al., 2020;Johnstone et al., 2010;Zheng et al., 2018).
In agreement with the above-mentioned findings on STIM1, optogenetic studies (He et al., 2015(He et al., , 2017;;Ishii et al., 2015;Kyung et al., 2015;Ma et al., 2020) proved the functional role of the CC1α1-CC3 clamp employing the homomerization domain Cry2 originating from Arabidopsis thaliana or the light-oxygen-voltage 2 (LOV2) domain from the phototropin of Avena sativa fused to a C-terminal fragment of STIM1.These light-responsive constructs remain in an inactive state in the dark and can be activated by blue light, either by oligomerization of Cry2 or by uncaging of LOV2.Cry2-STIM1-CT remained in the inactive state in the dark as long as the N-terminal CC1 residues after position 251 were present (Ma et al., 2020).LOV2-SOAR, co-expressed with STIM1 truncated after CC1 (STIM1 CC1) interacted with each other only under quiescent cell conditions and after blue light irradiation, which is consistent with the formation of the inhibitory clamp under quiescent conditions (Ma et al., 2020).
The effects of site-directed mutagenesis and helical wheel modelling led to the hypothesis that CC1α1 and CC3 are aligned antiparallel to each other, stabilized by the interacting pairs L258-V419 and L261-L416 (Fig. 3B) (Ma et al., 2015).Structural studies have so far not been able to resolve the CC1α1-CC3 interaction interface, since STIM1 structures available to date do not include CC1α1 and CC3 together.A recent single-molecule FRET (smFRET) study together with a lysine cross-linking approach confirmed the CC1-CC3 clamp and unmasked the interaction interface (van Dorp et al., 2021), however, suggesting a parallel orientation of CC1α1 and CC3 to each other.Direct interactions were supposed for L248 and L251 (CC1α1) with L416 and V419 (CC3), and for L258 and L261 (CC1α1) with L423 and L427 (CC3) (Fig. 3B) (van Dorp et al., 2021), in line with the activating effects of previously published point mutations (L248S, L251S, L416G, L258G/A, L261G, L423G) (Fahrner et al., 2014;Ma et al., 2015;Muik et al., 2011;Zhou et al., 2013).Moreover, STIM1 C-termini engage not only in an intramolecular 'cis'-CC1-CC3 coupling, but also switch between an intramolecular 'cis' and an intermolecular 'trans' conformation, which was proposed to enhance the stability of the inactive state and foster cooperation between subunits during structural transformations (van Dorp et al., 2021).
Additionally, smFRET uncovered CC1α2 and CC1α3 interactions with the binding interface formed by L286-A331, I290/A293-L328/V324, and A297/L300-L321/A317 (Fig. 3C) (van Dorp et al., 2021).Interestingly, this CC1α2-CC1α3 interface was directed away from CAD in the model based on the smFRET studies.While CC1α1/α2 is extended, in a recent NMR study, a three-helix bundle formed by the STIM1 CC1α1/α2/α3 helices was observed (Rathner et al., 2021).In this structure, the main interaction sites are M244-L321, L258/L261-I290/A293 and L248/L251-L300/L303 located at the CC1α1-CC1α2 interface (Fig. 3D).Weakening or disruption of these interactions (I290S/A293S; L300S/L303S) led to a reduced velocity and/or extent of STIM1/Orai1 activation and counteracted the effect of the Stormorken gain-of-function (GoF) mutation R304W, which is located at the very end of the CC1α2 helix.Thus, these CC1α1-CC1α2 interactions are thought to weaken the CC1α1-CC3 clamp (Rathner et al., 2021).These results further indicate that the GoF L251S mutation in STIM1 plays a role in mediating CC1α1-CC1α2 interactions (Grabmayr et al., 2020) in addition to its role in maintaining the inhibitory clamp and triggering STIM1 homomerization (see 'STIM dynamics after Ca 2+ store depletion').Interestingly, combining these CC1α2 mutations with STIM1 C227W or STIM1 D247C failed to affect their constitutive activity, suggesting distinct ways of action of the different GoF mutants J Physiol 602.8 (Grabmayr et al., 2020).Although this CC1α1/α2/α3 triple coiled-coil interaction was not visible in smFRET studies, the CC1α2-CC1α3 interaction appeared to be quite flexible.It was therefore suggested that the NMR structure might represent a preactivated state prior to full activation (Fig. 1) (van Dorp et al., 2021).Site-directed mutagenesis studies on key residues in CC1α3 that were proposed to mediate the interactions with CC1α2 (e.g.A331, L328, V324, L321, A317) may provide additional insight into their functional role in the STIM1 C-terminus, for instance to what extent they impact the CC1α1-CC3 clamp.
A recent alanine mutation scanning study has revealed that in addition to the previously identified critical hot spots at the CC1α1-CC3 inhibitory clamp interface, hydrophobic amino acids in CC2 (L347, W350, T354, H355, V357 and E358) and the apex region (L390, F391 and F394) of the CAD/SOAR domain enhance the tight conformation of STIM1 (Fig. 3E) (Shrestha et al., 2022).Thereby, critical residues identified in CC2 are located predominantly at the CC2-CC3 interaction interface observed in the human SOAR crystal structure (Yang et al., 2012) and in smFRET studies (Fig. 3E) (van Dorp et al., 2021).Interestingly, L373A in CC2 strengthened the CC1α1-CC3 clamp (Shrestha et al., 2022).Further screening by site-directed mutagenesis revealed a correlation between side chain size and the strength of the inhibitory clamp: the smaller the side chain (A, S, C, V) the stronger the CC1α1-CC3 clamp (Shrestha et al., 2022).We previously reported that L373S abolishes the direct STIM1-Orai1 interaction (Frischauf et al., 2009).These recent results suggest that our previously identified inhibitory effect of L373S on Orai1 activation occurs upstream of the STIM1-Orai1 coupling signalling step.Additional simulations that force the unfolding of CC1-CC3 and subsequent wet-lab experiments would be very interesting to potentially learn more on the mechanism of the inhibitory clamp within the entire STIM1 C-terminus.
A detailed characterization of F394 at the apex by site-directed mutagenesis revealed that its substitution can lead to changes along the CC1α1-CC3 interface, impact homo-oligomerization and impair coupling to Orai1 (Fig. 3E) (Hoglinger et al., 2021).Overall, these findings indicate that loss-of-function (LoF) mutations in the apex affect primarily the inhibitory clamp which then impedes downstream signalling steps.The observed effects on these different signalling steps are, however, still controversial.While in the STIM1-CC1-CAD FRET assay system, F394 substitutions weakened the CC1α1-CC3 clamp (Shrestha et al., 2022), incorporation of these mutants in the OASF-FRET sensor led to no changes in the inhibitory clamp (Hoglinger et al., 2021).One possible reason for these differences could be that either the OASF fragment, or the CAD fragment or even both adopt a less native conformation compared to their integration into full-length STIM1.In contrast, molecular dynamics simulations revealed a strengthening of the CC1α1-CC3 clamp upon F394 mutation (F394D/K), suggesting that additional studies are still required (Hoglinger et al., 2021).
Despite the discrepancy in the effects on CC1α1-CC3, the combined approach of functional, fluorescence and molecular dynamics simulation studies on the effects of LoF mutations in the apex (F394D/K) unveiled the orientation of CAD/SOAR within the entire STIM1 under resting cell conditions (Hoglinger et al., 2021).Until these studies, it remained unclear whether the CAD/SOAR apex is oriented toward or away from the ER membrane (Prakriya & Lewis, 2015).We discovered (Hoglinger et al., 2021), in line with the smFRET study of van Dorp et al. (2021), that the CAD/SOAR apex is in close proximity to the ER membrane under resting conditions.The LoF of STIM1 F394D/K could be explained in part by its tight attachment to the ER membrane.Whether lipid-apex interactions are involved in this inhibitory effect of the LoF mutations or even in the maintenance of the quiescent state of STIM1 wild-type still requires further investigations.
The inhibitory clamp is further supported by the inhibitory domain (aa 470−491) located downstream to CAD (Shrestha et al., 2022).In support, mutation-triggered complete release of CAD in the two-component FRET pair assay was impaired when CAD was fused to the inhibitory domain (aa 342−491).These findings are further in line with observations in full-length STIM1, where only some mutations found in the two-component FRET pair system were able to activate it, while others (e.g.L347A, T354A, E425A) did not, likely due to the presence of the ID (Shrestha et al., 2022).The different STIM1 isoforms all contain amino acid inserts of different lengths downstream of the ID.It would be interesting to investigate whether and how these inserts impact the CC1α1-CC3 clamp.
Despite the SOAR region of STIM1 and STIM2 being almost fully conserved, the CC1-CC3 interplay within STIM2 is much weaker.This observation adds to differences in the luminal STIM1 domain that results in enhanced clustering of STIM2 within the ER-PM contact sites observed already under resting conditions (Subedi et al., 2018).
This insightful work on the interplay of STIM1 C-terminal coiled-coil domains motivated us to model a STIM1 molecule comprising structurally resolved fragments (Fig. 3F, aa 63−443).In this model, we considered the configuration determined by van Dorp et al. ( 2021) and used the crystal structure of the SOAR domain (PDB: 3TEQ) (Yang et al., 2012).Moreover, we attempted to also add the TM domain and the NMR resolved luminal domain of STIM1 (PDB: 2K60) (Stathopulos et al., 2008).The Modeller software (Fiser & Sali, 2003) was employed to integrate the different subdomains into a reconstructed whole protein (Fig. 3F).We chose to model the TM domain and the CC1α1 domain as a continuous α-helix because we anticipated that only this configuration would facilitate communication from the luminal domain through the membrane and to the rest of STIM1 during activation.Any disruption or loose configuration, such as a coil, as opposed to a more structured configuration, could reduce or prevent potential rotation or translation from the TM domain due to the increased degrees of freedom.Once a monomer of STIM1 was created, SOAR was used to position the next monomer by aligning their Cα atoms.Subsequently, molecular dynamics simulations were performed to relax the structure.Interestingly, it was not possible to adjust the contact pairs as highlighted in van Dorp et al. ( 2021).In our model, the contact pairs had to be shifted to prevent a significant portion of SOAR from being at the level of the membrane.If the contact pairs had been respected, half of SOAR would be buried in the membrane, corresponding to about half of the TM helix.One way to reconcile our model with the experimental observations would be to consider an unstructured domain between the TM and CC1α1.This would allow increasing the distance of the CC1α1 domain to the membrane and fitting the contacts to CC3 as detected by smFRET (van Dorp et al., 2021), thus preventing SOAR from penetrating deeply into the hydrophobic core of the membrane.
A full atomic resolution of the entire STIM1 C-terminus in the quiescent conformation is highly awaited to understand the direct and allosteric interplay of the various coiled-coil regions as well as their configuration with respect to the ER membrane.Moreover, structural characterization of different key mutants within STIM2 compared to STIM1 is still required to understand their differences despite similar functional behaviour, e.g. for distinct GoF mutants.
STIM dynamics after Ca 2+ store depletion STIM N-terminus in the Ca 2+ -depleted state.Ca 2+ store depletion leads to structural alterations within the EF-SAM complex, which subsequently induces conformational changes in the C-terminus of STIM1.Initial investigations demonstrated, in the absence of Ca 2+ , a dramatic decrease in the α-helical content of the isolated EF-SAM domain.These studies also showed that the EF-SAM domains of STIM1 proteins form dimers and oligomers in the Ca 2+ free state, indicating that after the loss of Ca 2+ , structural changes and dimerization/oligomerization events are starting events in the STIM activation cascade (Furukawa et al., 2014;Stathopulos et al., 2006Stathopulos et al., , 2008)).Independent studies also confirmed oligomerization as a primary event for STIM activation by demonstrating increased FRET between luminal domains after store depletion, even in the absence of the C-terminal domain (Covington et al., 2010;Liou et al., 2007;Muik et al., 2008).By replacing the STIM luminal part with an FKPB12 and an FRP domain, respectively, and co-expressing these constructs, Luik et al. (2008) demonstrated, by chemically induced cross-linking approaches, that the association of the FKBP12 and FRB domains is able to activate STIM1.
Detailed observations over the years revealed that the highly stable and compactly folded EF-SAM structure present under Ca 2+ -bound conditions undergoes structural changes and oligomerization during Ca 2+ loss.Furukawa et al. (2014) showed that the EF-SAM domain cooperatively transits from a well-folded state to a mostly unfolded-unstructured state, and within the same study two crucial dimerization regions were identified.These are the SAM C-terminal helix D183-F200 and the SAM N-terminal helical hairpin V134-L157 in STIM1.Their molecular dynamics simulations suggested electrostatic repulsion between negatively charged residues in the EF-hands, which promote unfolding under Ca 2+ -reduced conditions.Additional NMR data strongly suggest a dynamic ensemble of conformational states during this unfolding process, consistent with circular dichroism results.The importance of the EF-hand-SAM in oligomerization processes was also supported by another study, demonstrating that the SAM domain displays a clear tendency to form dimers/oligomers in the apo (Ca 2+ free) form (Stathopulos et al., 2006).By analogy, optogenetics studies with Cry2 linked to N-terminal parts (EF-hand, SAM) of STIM1 revealed that in particular SAM is a key domain within the N-terminus for STIM1 oligomerization (Ma et al., 2020).Furthermore, mutational studies as well as STIM1 truncations (delta SAM-domain truncations) revealed the importance of the EF-SAM domain during oligomerization of human STIM1 and showed that the disruption of Ca 2+ binding, EF-hand-SAM interaction and SAM hydrophobic core mutations can induce STIM1 aggregation independently of ER Ca 2+ depletion.This suggests, as also shown by Furukawa et al. (2014), a SAM interaction mechanism that depends on the stability and folding of this domain, highlighting the importance of controlled Ca 2+ sensing by the EF-hand motifs and the essential stability of the hydrophobic core (Bohm et al., 2013;Sallinger et al., 2020;Schober et al., 2019a;Stathopulos et al., 2006Stathopulos et al., , 2008)).Precisely, loss of Ca 2+ binding triggers a more parallel helical conformation of EF-hand domains, which disrupts the EF-SAM interface and thus the hydrophobic cleft.Exposed hydrophobic residues cause an unstable conformational state that subsequently triggers oligomerization processes (Stathopulos et al., 2006(Stathopulos et al., , 2008)).The importance of structural stability in this J Physiol 602.8 oligomerization mechanism was further highlighted by Zheng et al. (2011), creating different sets of 'superstable' or 'unstable' STIM1-STIM2 EF-hand-SAM chimeras (super unstable: canonical EF-hand STIM2, non-canonical EF-hand STIM1, SAM-domain STIM1; super-stable: canonical EF-hand STIM1, non-canonical EF-hand STIM2, SAM-domain STIM2).In addition, these studies demonstrated stability differences between the isoform-specific domains of STIM1 and STIM2 proteins.Specifically, a chimera including the Ca 2+ -binding EF-hand of STIM1 and the SAM domain of STIM2 contains residues that have additive effects on the autoinhibition of oligomerization and act most efficiently compared to the other STIM isoform.Hence, although the EF-hand of STIM2 is activated with minimal store depletion, its SAM domain limits the activation process (Zhou et al., 2009).Structural differences were also investigated for the apo STIM2 EF-SAM domain.This showed that the STIM2 isoform is more stable, retaining most of its secondary structure in the Ca 2+ -unbound form.Size exclusion chromatography indicated that STIM2 EF-SAM exists as a monomer in both the Ca 2+ -bound and -depleted form at 4°C but undergoes aggregation at 25°C, in addition to seeming less compact (more extended) in molecular shape compared to the Ca 2+ -bound form (Stathopulos et al., 2006;Zheng et al., 2008).The first-mentioned feature is different from STIM1 EF-SAM, which shows aggregation even at 4°C.A recent paper by the Ambudkar laboratory gave further insights into the function and activation process of STIM2, demonstrating that endogenous STIM2 proteins are constitutively localized in immobile and mobile clusters at ER-PM junctions, which again highlights differences in STIM1-STIM2 protein functions (Ahmad et al., 2022).
Solution NMR investigations of the Ca 2+ -unbound state of Caenorhabditis elegans EF-SAM STIM luminal domain showed structural unfolding and α-helical to β-strand transition of the terminal SAM domain α10 helix, again indicative of the importance of this region after the loss of bound Ca 2+ (Enomoto et al., 2020).In comparison to the above-mentioned partial unfolding of STIM proteins, Gudlur et al. (2018) proposed, based on circular dichroism measurements using their EF-SAM-GrpE chimera, that nearly all α-helical secondary structures were retained in the absence of Ca 2+ and that large scale unfolding of the EF-SAM domain was not required to induce an active STIM1 conformation.A more detailed picture of the structural changes of Ca 2+ -depleted N-terminal domains of STIM proteins is thus still required.
STIM1 TM domain in the Ca 2+ -depleted state.Structural changes and oligomerization of the N-terminal part raised the question of how conformational rearrangements within the STIM1 N-terminus trigger structural changes in the TM domains that propagate to the C-terminus and lead to full STIM1 activation.NMR and luminescence resonance energy transfer (LRET) experiments unveiled distinct conformations of STIM1 TM domains in the quiescent state compared to the activated conditions.They showed that store depletion reduces the crossing angle of two interacting TM domains within a STIM1 dimer, decreasing the distance between the N-terminal segments (aa 214-220) (Ma et al., 2015).Consistent with these findings, cysteine cross-linking of the TM helices of two STIM1 proteins after store depletion resulted in a closer distance and a change in orientation with respect to each other.In fact, cross-linking of cysteines in the C-terminal portion of the TM domain (L216, S219, G223, G226, A230, Q233) occurred only in the active state (Fig. 3G) (Hirve et al., 2018;Ma et al., 2015).These structural changes essential to confer the activation signal from the ER luminal to the cytosolic side are likely facilitated by three glycines (G223, G225, G226) in the STIM1 TM domains (Dong et al., 2012;Ma et al., 2015).Moreover, these conformational rearrangements along the TM domain are associated with continuous coiled-coil formation from the TM domain to L251 in the proximal part of CC1α1 (Hirve et al., 2018).It is noteworthy that structural rearrangements of STIM1 TM segments not only are controlled by Ca 2+ in the ER but can also be manipulated by changes in the C-terminus, as demonstrated by the L251S mutation.This L251S substitution leads to constitutive activation of STIM1 and allows for a similar TM-cysteine cross-linking pattern, already in the presence of Ca 2+ , as observed in STIM1 wild-type in the absence of Ca 2+ (Hirve et al., 2018).These observations suggest that while the STIM1 activation signal is conferred from the N-to the C-terminus, structural rearrangements within STIM1 can occur bidirectionally.
STIM C-terminus in the Ca 2+ -depleted state.The next step in the STIM1 activation process is the signal transmission from the TM domain to the C-terminus.This leads, on the one hand, to a conformational change of STIM1 C-terminus from the tightly folded to an extended state and on the other hand to oligomerization of the cytosolic segment.The primary step in the elongation process is the loosening of the intramolecular clamp between CC1α1 and CC3, which results in the exposure of SOAR (Fahrner et al., 2014;Ma et al., 2015;Zhou et al., 2013).STIM1 fragments labelled on both sides exhibit a reduction in intramolecular FRET/LRET when the intramolecular clamp is released either by single point mutations such as L248S, L251S and L258S, by deletion of CC1 or by coupling to Orai1 (Muik et al., 2011;Zhou et al., 2013).Similarly, ER-bound CC1 regions cannot bind to CAD/SOAR fragments when mutated (e.g.L251S) (Fahrner et al., 2014;Ma et al., 2015).This approach together with an alanine scan confirmed known and identified new mutations (M244A, D247A, L248A, L251A, L258A) in CC1α1 leading to the release of SOAR (Shrestha et al., 2022).Moreover, CC1 fragments appear as monomers in solution and couple to CAD, whereas artificially cross-linked CC1 proteins form dimers that exhibit impaired coupling with cytosolic CAD/SOAR, consistent with di/homomerization of the CC1 domain and facilitating the switch to an elongated conformation (Zhou et al., 2013).
In agreement, Cry2-STIM1-CT activated already under dark conditions when CC1 was truncated upstream of aa 251.Cry2-STIM1-CT even enabled the identification of novel GoF and LoF mutations and the efficient characterization of the effects of disease-related mutations.Additionally, LOV2-SOAR and STIM1 truncated after CC1 (STIM1 CC1) did not couple to each other after store depletion and blue light irradiation, consistent with the release of the inhibitory clamp after store depletion (Ma et al., 2020).
Furthermore, CC1α2 was demonstrated to affect STIM1 function, as the deletion of this segment significantly slowed down STIM1 CC1α2-mediated Orai1 activation.In line, site-directed mutagenesis revealed that CC1α2 mutations (I290S+A293S, L300S+L303S) delayed store-operated activation.The reason for that represents a strengthened CC1α1-CC3 clamp (Fig. 3D).Whether mutations in CC1α3 may also modulate the strength of the inhibitory clamp still requires further investigation (Rathner et al., 2021).
Finally, subsequent to unfolding, the STIM1 C-terminus elongates and oligomerizes with adjacent C-termini.Disruption of the CC1-CC3 inhibitory clamp results in increased CC1 homomerization and SOAR exposure, whereas CC3 participates in higher order STIM1 oligomerization (Fahrner et al., 2014;Ma et al., 2015;Muik et al., 2009Muik et al., , 2011;;Stathopulos et al., 2013;Zhou et al., 2013).FIRE studies revealed most pronounced homomerization of CC1 and CC3, indicating their critical role in STIM1 oligomer formation.Moreover, single point mutations identified in full-length STIM1 to release (L251S) or strengthen (R426L) the inhibitory clamp showed corresponding effects in the FIRE system (Fahrner et al., 2014).Contrary to this, optogenetic studies performed with STIM1 fragments of different length attached to Cry2 revealed that their light-induced oligomerization necessitates only the presence of SOAR, while the presence of CC1 alone is dispensable (Ma et al., 2020).These different findings might be due to distinct fragment length and require additional investigation in full-length STIM1.Recent cysteine cross-linking studies in the quiescent compared to the activated conformation of STIM1 revealed a close pairing of the CC1 helices, with in particular A268 and T307 as contact points exhibiting much stronger cross-linking efficiency in the activated compared to the inactive conformation.S339 at the linkage region between CC1α3 and SOAR is already in a close contact point in the resting state and interestingly exhibited a slight decrease in distance after store depletion, indicating that the release of SOAR moves S339 positions a bit apart from each other.In support, cysteine cross-linking within STIM1 A268C and STIM1 T307C, but not STIM1 S339C, impeded STIM1 deactivation (Fig. 3G) (van Dorp et al., 2021).Overall, these findings indicate that a close pairing of the STIM1 C-termini is crucial for the efficient release of CAD towards Orai1.
The recently presented CC1-CAD model defines possible types of conformational changes required to transition STIM1 from its folded to its elongated conformation.Van Dorp et al. (2021) suggested two possible variants of structural alterations.First, CAD could undergo a symmetric conformational change in a folded-out model, with its apex getting removed from the ER membrane.In this case, one must assume that the CC2 domains transiently adopt an anti-parallel state similar to the NMR structure (Stathopulos et al., 2013).Interestingly, however, cysteine cross-linking results suggest that the C-termini of CC1α3 are close to each other in the activated state, but this does not agree with the spread configuration of the CC1α3 domains seen in the NMR structure.This antiparallel conformation is possibly present as an intermediate during the activation of STIM1.Second, there may be a flip-out mechanism whereby CAD would undergo an asymmetric outward rotation of 180°.However, smFRET data suggest that the release of CC1α1 does not cause a substantial conformational rearrangement of CAD.Overall, these findings reveal that STIM1 activation involves a drastic conformational rearrangement, whereas further studies are required to understand how all currently resolved structures, likely intermediate states, fit into the sequence of the activation cascade.In particular, to give a detailed atomic description of the STIM1 activation mechanism, the intermediate conformations of STIM1 still need to be resolved, and NMR and smFRET are promising strategies.
The STIM2 splice variant STIM2.1 is expected to show an altered configuration of the CC1-CAD interplay.STIM2.1 interacts with, but does not open, Orai1 and negatively impacts STIM1 and STIM2 function.This inhibitory effect was attributed to the 8 aa (VAASYLIQ) J Physiol 602.8 insert in the STIM2 CAD region, which likely impacts stable CAD formation (Miederer et al., 2015).Indeed, structural characterization of the OASF region carrying the VAASYLIQ showed that the overall α-helicity of OASF is reduced by the insert, which causes dysfunction.This results in disrupted coupling to the Orai1 C-terminal cytosolic helices, precluding Orai1 channel activation.Since it is known that both CC2 domains in a STIM dimer are essential to form the binding pocket that binds the Orai1 C-terminal cytosolic helices, heterodimerization with STIM1 and STIM2 is the most likely cause for STIM2.1-mediatedinhibition.Thereby, a structural change within one CC2 domain potentially causes the collapse of the whole binding pocket (Chung et al., 2018;Stathopulos et al., 2013).Downstream to the ID, STIM1 also contains the EB domain, which governs the attachment of STIM1 to growing tips of microtubules through the interaction with EB proteins (Grigoriev et al., 2008;Pchitskaya et al., 2017).This triggers the rapid movement of STIM1 proteins through the cell, although it seems to be dispensable because silencing of EB1 failed to block SOCE.Intriguingly, recent reports suggested the STIM1-EB1 coupling delays STIM1 migration to ER-PM contact sites and subsequent coupling to Orai1 (Chang et al., 2018;Wang et al., 2018).In this context, the detachment of STIM1 from microtubules was shown to be initiated by Ca 2+ store depletion (Pozo-Guisado et al., 2013;Sampieri et al., 2009).

Sites within STIM1 required for Orai1 interaction
Subsequent to STIM1 extension, CAD couples to Orai1 to induce CRAC entry into the cells.This event is initiated by the interaction of the STIM1 C-terminus with the Orai1 C-terminus, forming the main coupling site within the CRAC channel complex (Fig. 1).Small STIM1 C-terminal fragments (OASF, SOAR, CAD or Ccb9) (Kawasaki et al., 2009;Muik et al., 2009;Park et al., 2009;Yuan et al., 2009) identified to be sufficient for Orai1 activation contain all of CC2 (aa 345−391) and part of the extended CC3 (aa 393−450) region, which comprises the STIM1 homomerization domain (SHD, aa 421-450) (Muik et al., 2009).The SOAR/CAD region represents the minimal interaction partner for Orai1 sufficient to activate Ca 2+ entry.
In early attempts, CAD was shown to directly interact with Orai1 (Park et al., 2009).A search for STIM1 and Orai1 fragments competent for interaction revealed robust interaction between CAD and the Orai1 C-terminus, while only a weak interaction was identified with the Orai1 N-terminus and no interaction with the loop2 (Derler et al., 2013;Fahrner, Pandey et al., 2018;Park et al., 2009).In FIRE experiments, we discovered interaction with a STIM1 C-terminal fragment employing a slightly longer Orai1 loop2 fragment (Fahrner, Pandey et al., 2018).
Later functional and structural studies consistently revealed that store-operated STIM1-Orai1 coupling is initiated by the interaction of STIM1 C-terminus with Orai1 C-terminus, forming the main coupling site within the CRAC channel complex.A detailed resolution of this coupling site was achieved by the NMR structure of SOAP (STIM-Orai association pocket) showing the antiparallel association of two STIM1 C-terminal fragments (aa312-387) with two Orai C-termini (Fig. 1).This complex revealed that the residues L347 and L351 of one CC2, Y362, L373 and A376 of the second CC2 but also a positively charged cluster (K382, K384, K385 and K386) are crucial for the interaction (Stathopulos et al., 2013).These findings are in line with previous reports which demonstrated that mutations within SOAP (L347R, L351R, L373S, A376K) interfere with STIM1-Orai1 coupling (Frischauf et al., 2009;Stathopulos et al., 2013).However, it is still a matter of debate whether the SOAP binding pocket which has so far exclusively been resolved with fragments also occurs in the full-length CRAC channel complex.Noteworthy, the resolved STIM1 SOAP structure was unable to bind and activate Orai1 in functional studies (Muik et al., 2009) in the absence of CC3.Alternatively, this SOAP structure could represent an intermediate binding conformation.
Within SOAR, the Sα2 segment was proposed to contribute to the coupling and activation of the STIM1-Orai1 complex.Initial studies pointed to a possible role of F394 in mediating the interaction with Orai1, due to a reduced function or LoF of several mutations within this position.Isoform specific differences in the context of this residue and affected function of STIM1 F394H could be restored in a heterodimer with STIM1 wild-type (Wang et al., 2014b;Zhou et al., 2015;Zhou et al., 2018).However, some studies have recently shown that the effects of LoF mutations at F394 tend to underlie processes upstream of STIM1-Orai1 coupling, as described in 'Key features of STIM in the resting state' (Hoglinger et al., 2021).Independent of the effects of mutations on F394, it is still probable that the apex in STIM1-WT functions as a binding site for Orai1.
C-terminal to Sα2 in SOAR, Sα3 was reported to be critical for CRAC channel activation, by contributing to the transmission of the STIM1-induced activation signal to the pore (Butorac et al., 2019).Thereby, STIM1 Sα3 deletion or point mutants can still bind to but cannot activate Orai1.Cysteine cross-linking allowed for the identification of critical positions involved in the functionally relevant interaction of Sα3 with Orai1.In particular, L402 in STIM1 and E166 in Orai1-loop2 were identified as important positions.Electrophysiological studies showed that cross-linking of STIM1 L402C and Orai1 E166C activated strong CRAC currents, independently of STIM1.This highlights the functional relevance of this interaction, named the STIM1-Orai1 gating interface (SOGI) (Butorac et al., 2019).The study of Sα2 and Sα3 in SOAR yielded important insights which determine the molecular interplay within the CRAC channel complex.Further in-depth structural resolution of the CRAC channel complex is still required to substantiate the interactions proposed from the results of the functional studies.
smFRET studies and cysteine cross-linking studies of CAD revealed major differences compared to the crystal structure in the apex, comprising the Sα2 and Sα3 helices.The apex appears more flexible and residues located opposite are farther apart in the smFRET and cross-linking approaches than in the crystal structure.The flexibility of the CAD apex is of interest because site-directed mutagenesis and cysteine cross-linking studies propose that this segment couples to Orai1 to trigger channel activation (Butorac et al., 2019;Calloway et al., 2010;Korzeniowski et al., 2010;Thompson et al., 2018;Wang et al., 2014b).Conformational flexibility may ease protein-protein interactions through impacts on binding thermodynamics (Grunberg et al., 2006).Consequently, it is possible that the apex adopts a different structure in the STIM1-Orai1 complex than in previous structural analyses.
STIM1A, containing the 31 aa-long domain A downstream of the ID, decreases CRAC currents in a dominant-negative manner.A conserved phenylalanine-serine-aspartic acid motif was identified to be critical for STIM1A function, as mutations (S502A, D503A) therein gave rise to functions reminiscent of STIM1.Thus, it was hypothesized that the C-terminal insert in STIM1A perturbs a stabilizing interaction between the STIM1 CAD/SOAR region and the extended transmembrane Orai1 N-terminal (ETON) region, which is a prerequisite for full Orai1 gating (Knapp et al., 2022).It is conceivable that the alanine substitutions trigger conformational alterations making CAD/SOAR better accessible for coupling to Orai1.
Collectively, SOAR within STIM1 C-terminus contains the major binding site for Orai.Whereas the coupling of the C-termini of STIM1 and Orai1 is clearly defined, the amino acids which are involved in potential interactions with the other cytosolic regions of Orai channels are presently not clearly defined or unknown.Furthermore, a detailed characterization of the role of sequence alterations in STIM isoforms in the coupling to Orai would provide novel mechanistic insights.

Diseases associated with STIM isoforms or dysregulation
Proper functioning of STIM proteins involves concerted and extensive conformational adaptations and depends on the Ca 2+ -filling state of the ER, as detailed previously.A series of alterations within STIM1 were already described to interfere with this precise choreography, causing either constitutive activity or preventing Ca 2+ entry altogether when stores get depleted.Thereby, several GoF mutations were already proven as causal for three rare, phenotypically overlapping conditions: the progressive muscle disorder tubular aggregate myopathy (TAM), Stormorken syndrome and York platelet syndrome (YPS) (Markello et al., 2015;Riva et al., 2022).TAM, which may further result from mutations in other proteins like Orai1 or the Ca 2+ -buffering protein calsequestrin-1 (Barone et al., 2017;Feske, 2019;Nesin et al., 2014), shows some clinical heterogeneity and may involve myasthenic features or weakness of proximal muscle, cramps and muscle pain (Bohm et al., 2014).In any instance, however, muscle biopsies from TAM patients typically display densely packed membrane tubules which are eponymous for the overall condition (Bohm & Laporte, 2018;Ticci et al., 2021).There is considerable phenotypic overlap in YPS and Stormorken syndrome, whereby TAM is often part of a wider spectrum of clinical manifestations of both.Regarding the last-mentioned syndrome, thus far reported cases show TAM going along with hyposplenia or asplenia, bleeding diathesis, mild anaemia, short stature, ichthyosis, hypocalcaemia, miosis, dyslexia and low body weight (Claeys et al., 2020).Yet, given the mentioned reminiscence of the symptoms and especially also the fact that identical mutations were reported to give rise to both conditions, the dissection of YPS and Stormorken syndrome into two different diseases is questioned.Singh et al. (2015) raised in this regard concerns that assigning patients to YPS and Stormorken syndrome based on the apparent presence or absence of specific phenotypic expressions may simply be flawed on account of the methods actually used to study clinical specimens and suggested combining both in one spectrum disorder, Stormorken-York platelet syndrome.Despite such reasonable concerns, we will herein mention whether GoF mutations were associated with a particular syndrome, as most of the publications wherein these have initially been described or further characterized follow the classical distinction.LoF mutations also have consequences for skeletal muscle, such as muscular hypotonia, and lead in addition to immunodeficiency with impaired functions of T lymphocytes, B cells and natural killer cells, predisposing patients to recurrent and chronic infections with bacterial, fungal and viral pathogens (Lacruz & Feske, 2015).Moreover, autoimmunity, severe defects in dental enamel formation and anhidrosis are phenotypic expressions of impairments in CRAC channel activation (Grabmayr et al., 2020;Lacruz & Feske, 2015).
Most of the disease-relevant STIM1 mutations reported to date are located in the EF-hand domains (e.g.H72Q, N80T, G81D/N, D84E/G, S88G, L92V, L96V, Y98C, L303P, J Physiol 602.8 K104N, F108I/L, H109N/R/Y, I115F), the majority of which were identified only in single families (Fig. 4A) (Bohm et al., 2013(Bohm et al., , 2014;;Claeys et al., 2020;Conte et al., 2021;de la Fuente-Munoz et al., 2022;Harris et al., 2017;Noury et al., 2017;Riva et al., 2022;Walter et al., 2015).A recent study also identified a variation within the SAM domain, V138I, which leads to an increase in STIM1 activity, whereas another disease-relevant mutation within the same domain, P165Q, has an opposite effect on protein function (Lacruz & Feske, 2015;Ticci et al., 2021).Although mutations within the TM domain were, as mentioned in a previous section, proven to be competent in triggering store-independent activation, no alterations within this region were thus far reported to be causal for disease.On the cytosolic side, the STIM1 E255V variant was supposed to have a pathogenic character but its consequences for protein function remain to be investigated more closely (Gang et al., 2022).Yet, the positioning close to CC1α1 residues which are essential for the resting-state CC1-CC3 clamp and the considerable change in the chemical properties makes it tempting to speculate that the mutation affects formation of the inhibitory clamp.However, it needs to be mentioned that no significant change in CC1-CAD binding was seen in a recent study by Shrestha et al. (2022) upon introducing an E255A substitution.Farther downstream in the C-terminus, substitutions like L303P, R304Q/W/G, K365N, S630F and R749H alter STIM1 function in a clinically relevant manner, as do the LoF alterations R426C and R429C, among others (Fig. 4B) (Bohm & Laporte, 2018;de la Fuente-Munoz et al., 2022;Harris et al., 2017;Morin et al., 2014;Riva et al., 2022).In the following sections, some mechanistic insights on mutation-based alterations in protein function will be reviewed.
N-terminus.Mutations within the N-terminus are prone to lead to store-independent activation of STIM1 if they interfere with Ca 2+ binding and/or folding (Fig. 4A).Indeed, substitutions altering one of the series of residues that contribute to the hydrophobic pocket responsible for packing upon rest were shown to be sufficient for persistent clustering of STIM1 and Ca 2+ entry.These GoF mutations include H72Q, L96V within the canonical domain and F108I, H109N, H109R, or I115F, each assigned to the non-canonical EF-hand domain (Bohm et al., 2013(Bohm et al., , 2014)).In the wild-type protein, these residues are compactly packed together under resting conditions to keep the luminal domain folded.Instead, in the absence of Ca 2+ , the conformation opens with F108 and H109 being released from the preserved remnants of the core of the hydrophobic pocket.Interestingly, comparison of Ca 2+ -free wild-type STIM1 and the F108I mutant in molecular dynamics simulations revealed differences in the positions undergoing the largest degree of unfolding, indicating that the mutation-based activated state does not perfectly mimic STIM1 after store depletion (Sallinger et al., 2020).Altogether, molecular dynamics simulations of Schober et al. (2019a)   destabilization.Moreover, mutations which reduce the structural stability may have consequences for Ca 2+ binding as well, whereby a significantly higher K D value compared to the wild-type was for instance determined for the F108I variant.Taken together, mutations within the canonical domain seem to be predestined to interfere with the maintenance of the hydrophobic cleft, leading to loss of Ca 2+ and in consequence promote dimerization of the luminal domains and STIM1 activation.Constitutively active STIM1 variants harbouring mutations within the non-canonical EF-hand domain were recently further shown to stimulate autophagy-associated processes, such as activation of the transcription factors MITF or TFEB and fostering the formation of autophagosomes (Sallinger et al., 2020).The STIM1 I115F mutation was reported to be competent for giving rise to a broader clinical spectrum than TAM, expressing in York platelet syndrome (Markello et al., 2015).Various GoF mutations of the hydrophobic cleft residues were also identified in cancer patients, including H72R and L92P, whereby the former mutation apparently reduces binding of Ca 2+ to the STIM1 D76-D78 region.Indeed, in the wild-type state, the association of a single Ca 2+ ion with D76 and D78 was shown to suffice for preserving the folded state, while extensive unfolding of the non-canonical EF-hand segment of STIM1 H72R is observable independent of the presence of Ca 2+ (Schober et al., 2019a).Considering the canonical EF-hand residues L92 and L96 whereof substitutions by the helix-breaker proline or valine, respectively, in TAM or in cancer tissue, these experience the most pronounced reorientation when wild-type STIM1 transits from the resting to the activated form (Sallinger et al., 2020).Indeed, in molecular dynamics simulations on the L92P mutant, extensive unfolding was detected of the non-canonical EF-hand rather than of the mutation-carrying canonical EF-hand domain itself.Other cancer or TAM related mutations are directly associated with the Ca 2+ binding loop of the canonical EF-hand domain, such as D76V, D78G, A79T, N80K, D84G and E87Q (Schober et al., 2019a).Within this location, removal of a negative charge or introduction of a positive charge is prone to interfere with Ca 2+ sensing and hinder engagement of a quiescent state if stores are full.Simulations of the E87Q mutant, which was indeed shown to have a lower affinity for Ca 2+ in UV-circular dichroism spectral analysis, indicated that the N-terminus gets destabilized due to the mutation yet not to the same level as wild-type STIM1 upon Ca 2+ depletion.Thereby, the mutation apparently triggers partial unfolding of the non-canonical EF-hand domain while the canonical counterpart remained stable during the entire simulation time.For STIM1 N80K, the non-canonical EF-hand was reported to unfold extensively, similar to the H72R mutant.In fact, some of the aforementioned mutants were proven to maximally interact with Orai1 already in ER Ca 2+ -replete conditions with no increase after store depletion.Consistent with this GoF, nuclear factor of activated T cells (NFAT) proteins were forced to accumulate within the nucleus without external stimulation (Schober et al., 2019a).The close-by mutations S88G and G81N, which affect conserved residues and were identified upon whole exome sequencing of YPS patients, are likely pathogenic as well, as predicted by in silico analysis, but experimental studies are still awaited (Harris et al., 2017).Apart from the syndromes which are conventionally associated with STIM mutations and cancer, the STIM1 E152K mutant was identified by Burgos et al. (2021) in three pancreatitis patients.The mutation was shown to promote Ca 2+ release from the ER in fibroblasts derived from patients and in the HEK293T expression system and an increase in SOCE.
The P165Q mutation within the STIM1-SAM does not result in GoF but stalls STIM1 activation by eventually interfering with dimer formation of EF-SAM domains (Lacruz & Feske, 2015).In other patients suffering from immunodeficiency, haemolytic anaemia, muscular hypotonia and thrombocytopenia, among others, an adenine insertion in exon 3 of STIM1 was detected, resulting in a scrambled sequence of eight amino acids followed by a premature termination codon (E136X).Considering that affected individuals show reduced STIM1 mRNA levels while the protein is undetectable overall, the resultant transcript seems to be degraded by nonsense-mediated mRNA decay (Picard et al., 2009).Interestingly, E136X was also identified in Stormorken syndrome patients, which is typically caused by a GoF mutation within the C-terminus of STIM1 (Huang et al., 2020).A splice-site mutation led to a quantitative deficiency, specifically a guanine to adenine substitution at the −1 position of exon 8 of STIM1.This alteration was identified in a child with classic Kaposi sarcoma a rare angiogenic, inflammatory neoplasm which may develop in particular in patients with immunodeficiency or immunosuppression in consequence of infections with human herpesvirus-8 (Byun et al., 2010).
C-terminus.Alterations within the cytosolic domain that endow STIM1 with store-independent activity include R304W, which was also extensively characterized in mice, or the YPS-linked R304Q (Fig. 4B) (Borsani et al., 2018;Fahrner, Stadlbauer et al., 2018;Gamage et al., 2018Gamage et al., , 2020;;Harris et al., 2017;Markello et al., 2015;Misceo et al., 2014;Morin et al., 2014;Nesin et al., 2014;Rathner et al., 2021;Silva-Rojas et al., 2021).Although typically associated with Stormorken syndrome, the R304W mutation is linked to a YPS diagnosis when patients show accompanying myopathy but no evidence J Physiol 602.8 of further characteristics of Stormorken syndrome (Harris et al., 2017).Interestingly, YPS cases caused by mutations in the STIM1 C-terminus and the N-terminal domain differ with respect to phenotypic expression and the age of onset, whereby previously mentioned alterations in the STIM1 N-terminus such as G81N, S88G or D84E are associated with additional and more pronounced clinical manifestations, such as contractures and progressing muscle weakness, and are likely to occur early rather than late (Harris et al., 2017).
Residue R304 is the one showing altogether the highest frequency of STIM1 mutations in patients (Bohm & Laporte, 2018).Mechanistically, constitutive activity of STIM1 R304W seems to be elicited in two ways.On the one hand, measurements with a FRET-based conformational sensor, biochemical efforts and NMR experiments collectively point to a localized helical elongation of the linker connecting CC1α2 and CC1α3.On the other hand, CC1 stretches of this mutant show a higher propensity to associate with each other, which together with the former, leads to a destabilization of the resting conformation, and the α2 helix is induced to elongate by about 3.5 additional helical turns toward the C-terminal region.Altogether, this culminates in the exposure of the CAD/SOAR domain to interact with Orai1 (Fahrner et al., 2014;Fahrner, Stadlbauer et al., 2018;Rathner et al., 2021).Interestingly, using the FIRE approach to probe CC1-CC3 interactions, the CC1-CC3 clamp interaction, which is vital for the quiescent state of wild-type STIM1, was shown to be per se not perturbed by the R304W mutation but is instead hindered due to steric reasons in the full-length protein (Fahrner et al., 2014;Fahrner, Stadlbauer et al., 2018).Furthermore, studies recently published together with the NMR solution structure of STIM1 CC1 wild-type suggested that R304W leads to a less compact structure with fewer interhelical contacts.In fact, store-dependency of full-length STIM1 R304W was shown to be restored by co-introducing I290S together with A293S or L300S + L303S into the Stormorken construct (Rathner et al., 2021).There are contradictory reports on the functional consequences of a missense/frameshift mutation combination in the cytosolic inactivation domain, STIM1 I484RfsX21 (Fig. 4B).Upon its initial characterization, this STIM1 variant identified in a TAM patient was reported to lead to a reduction in SOCE when expressed in HEK293 cells, contradicting the fact that TAM normally results from GoF mutations (Okuma et al., 2016).Recently, however, Kim et al. (2022) detected clustering, constitutive Ca 2+ influx and increased SOCE when the corresponding mutant was co-expressed with Orai1.In addition, the authors found that the mutation resulted in currents lacking SCDI, while FCDI was not altered.Another frameshift mutation was identified in a study of inherited myopathies as well, H632fs, with the consequences for protein function remaining to be clarified (Fig. 4B) (Ticci et al., 2021).
Two mutations within the cytosolic portion that are well documented to hinder activation are STIM1 R426C and R429C, both located within the CAD domain (Fig. 4B) (Fuchs et al., 2012;Maus et al., 2015;Wang et al., 2014a).Although eventual consequences of the R426C mutation for protein expression levels remain to be investigated, Shrestha et al. (2022) recently showed that R426C or R426A substitutions within CC3 hinder interactions with CC1α1, releasing the CC1-CC3 clamp.While this in itself would indicate a GoF, R426C concomitantly reduces the ability to interact with Orai1.This hindrance was found to be moderate yet significant for isolated CAD and more pronounced for full-length STIM1, accounting for a reduction in current activation (Shrestha et al., 2022).The relevance of R426 for the STIM1-Orai1 interplay was identified by Muik et al. (2011) already in a rather early study, showing that promotion of coiled-coil interactions due to R426L impedes interactions of both isolated OASF segments and full-length STIM1 with Orai1.
However, while the phenotypic consequences of R426C are limited to hypomaturation enamel defects and nail dysplasia rather than involving also skeletal muscle and the immune system (Wang et al., 2014a), the close-by R429C LoF mutation is expressed in a broader clinical phenotype including also immunodeficiency yet again seems to affect protein function in multiple ways (Fuchs et al., 2012;Maus et al., 2015).On the one hand, R429 is apparently essential for CC3 to engage a stable α-helical fold, considering that upon substitution of the residue by A, L or the disease-relevant C, far-UV circular dichroism spectra of isolated STIM1 C-termini showed an 8−9% lower α-helicity.Furthermore, thermal melt curve analysis pointed to a lower stability of the mutant proteins.Such distortions in protein structure were proposed to hinder CC3-related intramolecular interactions, as well as oligomerization of cytosolic STIM1 portions, although CAD dimerization remained intact in the study in question (Maus et al., 2015).The mutation apparently interferes with the association with Orai1, such that expression of STIM1 R429C/E or -L mutants together with Orai1 in HEK293 cells results in a reduction of SOCE below endogenous levels, although the extension of the STIM1 C-terminus led to constitutive localization at ER-PM junctions due to the exposure of the PBD.Indeed, co-expression of the aforementioned R429 variants with wild-type STIM1 indicates that these mutations impose a dominant negative effect (Maus et al., 2015).
Apart from such missense mutations or shifts in the reading frame and the occurrence of premature termination codons, changes in expression levels (Cheng et al., 2012) and/or the cell type-specific expression patterns of the different STIM1 isoforms have clinical relevance alike.Decreased STIM1 expression was linked to the development of combined immune deficiency, where patients often suffer from life-threatening infections (Vaeth et al., 2020).Deficiency in STIM1 and STIM2 in T lymphocytes was associated with the development of the exocrine gland autoimmune disease, Sjögren's syndrome.Enhanced STIM1 expression was recently reported for instance in myocardial specimens of heart failure patients compared to controls, leading, together with the concomitant reduction in STIM2.1, to an increase in SOCE (Cendula et al., 2019).During sepsis, STIM1 expression is increased in endothelial cells, which contributes to hyperpermeability (DebRoy et al., 2014).Diminished occurrence of STIM1 led to neurodegeneration, which is likely the cause for Alzheimer's disease (Pascual-Caro et al., 2018).Moreover, a reduced presence of the neuron specific STIM1 splice variant STIM1B, which is targeted to the presynaptic ER and functionally deviates from conventional STIM1, was associated with Alzheimer's disease (Ramesh et al., 2021).In contrast, expression of the muscle specific isoform STIM1L was reported to be elevated in hypertrophic right ventricular cardiomyocytes in a rat model while classical STIM1 was reduced (Sabourin et al., 2018).
Tumour development, aggressiveness and poor prognosis of a wide variety of cancers is often also associated with an altered expression of STIM1 and STIM2, as described in detail in Tiffner et al. (2022).For example, gastric and colorectal cancer tissues exhibit increased STIM1 expression in comparison with the respective non-cancerous tissues, which apparently promotes invasion and metastasis (Wang et al., 2015;Wen et al., 2016;Xia et al., 2016;Xu et al., 2016).Indeed, treatment with STIM1-targeted short hairpin RNA was shown to inhibit the migratory abilities of colorectal carcinoma cells.The increase in migration of colorectal cancer cells seems to be linked to an increase in cyclooxygenase-2 expression levels and prostaglandin E2 generation accompanied by aberrant STIM1 expression.STIM1 overexpression was, in the case of colorectal cancer, further reported to correlate with tumour size, among other things (Wang et al., 2015;Yang et al., 2017).Increased STIM1 expression levels are apparently also a critical determinant for breast cancer cells to migrate (Yang et al., 2009) and for the migration of cervical cancer cells, as well as for their proliferation and angiogenesis (Chen et al., 2011(Chen et al., , 2013)).Furthermore, STIM1 overexpression is crucial in the survival and migration of osteosarcoma cells (Zang et al., 2019).Recently, Wang et al. (2019) found STIM1 to be upregulated within the hypoxic microenvironment of pancreatic duct adenocarcinoma, which is based on enhanced transcription due to binding of the transcription factor hypoxia-inducible factor-1 to the STIM1 promotor.STIM1 and STIM2 levels were also found to be elevated in the case of diabetic kidney disease and were supposed to promote the disease-associated epithelial-to-mesenchymal transition of podocytes (Jin et al., 2019).

Additional factors modulating STIM function
Store-operated STIM1 activation triggers its association with Orai1 at ER-PM junctions (Zhou et al., 2010).Despite this interaction being sufficient for CRAC channels to open, their activity is modulated by a number of cellular components (Cao et al., 2015;Hogan, 2015;Maltan et al., 2022), such as directly by lipids (Balla, 2018;Bohorquez-Hernandez et al., 2017;Carreras-Sureda et al., 2021;Derler et al., 2016;Gwozdz et al., 2012;Pacheco et al., 2016), post-translational modifications or indirectly by lipid-dependent and ER-PM contact site-dependent accessory proteins (Berlansky et al., 2021;Son et al., 2016), as outlined below.In addition, it is currently emerging that STIM expression is regulated by transcription factors, which, although not a direct regulatory mechanism, plays an important role in identifying developmental stages and physiological as well as pathophysiological conditions in which STIM proteins are essential.Current knowledge of this form of modulation is summarized by Niemeyer (2016).
Phosphoinositides.Soon after the identification of STIM proteins, PIP 2 was shown to be pivotal for STIM1 function.Indeed, inositol 5-phosphatase, which decreases PIP 2 levels, reduced SOCE (Calloway et al., 2011;Korzeniowski et al., 2009).PIP 2 sensitivity of STIM proteins underlies a lysine-rich region at the very end of their C-termini (Collins & Meyer, 2011;Ercan et al., 2009;Heo et al., 2006;Liou et al., 2007;Park et al., 2009).With the latter, both STIM1 and STIM2 attach upon elongation to PM-PIP 2 and PM-PIP 3 in particular enriched in cholesterol-rich regions (Cao et al., 2015;Chvanov et al., 2008;Korzeniowski et al., 2009;Liou et al., 2007;Maleth et al., 2014;Walsh et al., 2010).Deletion or mutation of this polybasic region intervenes with the stable, store-operated coupling of STIM1 to the PM, despite STIM1 maintaining the ability to oligomerize (Liou et al., 2007).Thus, the lysine-rich region enhances the efficacy of STIM1 coupling to Orai1.Noticeably, STIM1 binding to PIP 2 necessitates the tetramerization of its polybasic segment, while efficient PIP 2 binding to STIM2 requires the dimerization of the lysine-rich J Physiol 602.8 region.This enhances the affinity for PIP 2 and lowers the activation barrier of STIM2 (Bhardwaj et al., 2013).Mechanistically, store-dependent STIM1 activation leads first to the assembly with PM-localized PIP 2 and PIP 3 in cholesterol-rich regions and in consequence to the interaction with Orai1 (Wu et al., 2014).This suggests that cholesterol indirectly drives the accurate STIM1-mediated targeting of SOCE components in lipid rafts by retaining the necessary phosphoinositide pools.
There is further evidence that the precursor of PIP 2 , phosphatidylinositol 4-phosphate (PI4P), modulates the function of CRAC channels (Broad et al., 2001;Korzeniowski et al., 2009;Rosado & Sage, 2000).Further studies and tools are required to clarify the direct or indirect role of PI4P on STIM1 (Bojjireddy et al., 2014).
Cholesterol.Cholesterol was demonstrated to modulate STIM1 function via direct interaction (Bohorquez-Hernandez et al., 2017;Derler et al., 2016;Pacheco et al., 2016).Chemical cholesterol depletion was demonstrated to enhance CRAC entry of STIM1 and Orai1 co-expressing cells (Derler et al., 2016;Pacheco et al., 2016).In search of potential cholesterol binding motifs, Pacheco et al. (2016) reported a well-known cholesterol binding motif with the consensus sequence L/V-X(1−5)-Y-X(1−5)-R/K in the C-terminus of STIM1.Indeed, the single point mutation, I364A, in this binding motif of both full-length STIM1 and a C-terminal STIM1 fragment led to a comparable increase in Orai1 currents to chemical cholesterol depletion.Molecular dynamics simulations confirmed that cholesterol impacts SOAR coupling to the membrane with I364 representing the major player at this interface.Additional application of methyl-β-cyclodextrin (MβCD) did not further enhance STIM1 I364A mutant-mediated Orai1 currents indicating that interference with the cholesterol-binding motif is sufficient to increase STIM1-mediated Ca 2+ entry via Orai1 (Pacheco et al., 2016).It is noteworthy that within the CRAC channel complex, not only STIM1 but also Orai1 contains a cholesterol binding motif (Derler et al., 2016;Maltan et al., 2022) the disruption of which results in increased Ca 2+ entry.Nevertheless, the effect of the removal of both binding sites is not additive, demonstrating that removal of one of the two binding sites is sufficient to mimic the effect of cholesterol removal from the membrane (Pacheco et al., 2016).Despite direct cholesterol binding being resolved for STIM1, several questions still remain open.While MβCD-mediated cholesterol depletion enhanced CRAC currents of STIM1/Orai1 overexpressing cells (Gwozdz et al., 2012), endogenous CRAC currents were reduced (Galan et al., 2010).Moreover, the application of MβCD was on the one hand shown to promote STIM1-Orai1 coupling, while on the other hand treatment with cholesterol oxidase or filipin left STIM1-Orai1 coupling unaffected (Derler et al., 2016).It is probable that the overexpression versus endogenous proteins leads to the distinct effects (Gwozdz et al., 2012).Alternatively, the observed differences arise potentially due to the harsher manipulation by MβCD than cholesterol oxidase or filipin.The latter potentially leave the membrane integrity intact or the distinct ways of chemical cholesterol depletion lead to differences in membrane composition.The relationship between STIM1 and cholesterol is still controversial in regard to its dynamics and physiological significance.A major question is also the timing of when cholesterol binding to STIM1 is present.Cholesterol could bind as early as when STIM1 couples to PM-PIP 2 or when it has formed a complex with Orai1.Cholesterol-dependent regulation of CRAC channel components may potentially have underestimated pathophysiological relevance.We hypothesized a link between hypocholesterolaemia and augmented mast cell degranulation.In fact, patients suffering from hypocholesterolaemia are prone to an elevated allergic reaction (Kovarova et al., 2006), consistent with the findings that cholesterol depletion in rat basophilic leukemia (RBL) mast cells potentiates store-operated Ca 2+ currents and degranulation (Derler et al., 2016).However, additional studies are required in this respect as cholesterol may impact several other mast cell signalling pathways.
Despite an available set of different STIM1 fragments, no structural information showing lipid-STIM1 interaction is available so far, but this is highly awaited to better understand the modulatory role of ER-PM junction lipids.

Modulatory role of post-translational modifications.
Various protein modifications were described to own a modulatory function on STIM activity, such as phosphorylation and acylation, whereby the latter links to the lipidic regulation.
STIM proteins can get phosphorylated by protein kinase C (PKC), with functionally relevant phosphorylation sites in the STIM1 N-terminus being the serine residues S27 and S30.Since mutations thereof that impede phosphorylation were shown to enhance CRAC channel activation, it is reasonable to conclude that PKC impairs CRAC channel activation through phosphorylation of these residues (Kawasaki et al., 2010).Furthermore, position Y361 in CC2, was reported to be phosphorylated by proline-rich kinase 2 after store depletion.Single point mutation of Y361F triggered STIM1 clustering, but coupling to Orai1 was impaired (Yazbeck et al., 2017).Additionally, STIM1 function is controlled by extracellular signal-regulated kinase 1 and 2 (ERK1/2), which triggers phosphorylation of residues in the serine/proline-rich region at the C-terminal end.This occurs during store depletion, is required for dissociation from EB1 and is recovered upon resetting STIM1 to the resting state (Pozo-Guisado et al., 2013).
S-acylation, also known as S-palmitoylation, refers to the reversible, post-translational linkage of medium-length fatty acids to a cysteine.This enables dynamic and efficient control of protein function and interaction (Chen et al., 2021;Kordyukova et al., 2019;Shipston, 2011).
STIM1 is also rapidly S-acylated to C437 after depletion of ER Ca 2+ stores.Indeed, a STIM1 mutant C437 could not be S-acylated, and this protein modification was shown to be essential for STIM1-Orai1 assembly into puncta and maximal CRAC channel activation (Kodakandla et al., 2022).
In addition to S-acylation, the cysteine's thiol group serves as reactive group for the linkage of other moieties.In this regard, Zhu et al. (2018) reported that S-nitrosylation of C49 and C56 increases the stability of the STIM1 N-terminus, suppressing hydrophobic exposure and oligomerization after Ca 2+ store depletion (Gui et al., 2018).In support, point mutation of C56 and S-gluthionylation reduced the Ca 2+ binding affinity of STIM1 and triggered structural deviations in the STIM1 EF-SAM complex (Sirko et al., 2022).
Moreover, glycosylation sites within the SAM domain control STIM1 function.In particular, a STIM1 N131D N171Q mutation leads to GoF, allowing faster Orai channel activation.Further investigations are required to understand how these mutations facilitate the transition of STIM1 from the closed to the open state.Intriguingly, this STIM1 DQ mutant decreased the level of Orai protein in the cell, indicating that an alternative mechanism causes enhanced Orai activation, particularly due to oligomerization rates (Kilch et al., 2013).

Direct modulation at the ER-PM contact sites
Junctate and junctophilin-4.Junctate is similar to STIM1, a Ca 2+ -sensing TM protein situated in the ER membrane.It binds directly to STIM1 and supports its store-operated clustering at the PM in a PIP 2 -dependent manner.However, the binding site within STIM1 is still unknown.
Truncation of the Ca 2+ sensing region of junctate ameliorated STIM1 puncta formation.Overall, it is assumed that junctate supports STIM1 in its movement to the ER-PM junctions and its consequent coupling to Orai1 (Srikanth et al., 2012;Treves et al., 2004).
Junctophilin 4, another junctional protein, also facilitates recruitment of STIM1 to ER-PM contact sites and STIM1-Orai1 coupling in a PIP 2 -independent manner.It was demonstrated to physically interact with STIM1 via its cytosolic domain, since a cytosolic fragment of junctophilin-4 interfered with SOCE.Furthermore, junctate and junctophilin-4 can form a complex that synergistically couples to STIM1 to support its accumulation at ER-PM junctions (Woo et al., 2016).
SARAF.SARAF constitutes a single TM protein residing in the ER, with the N-terminus directed toward the ER lumen and the C-terminus situated in the cytosol.Under resting cell conditions, it associates already with STIM1 at ER-PM contact sites to impede spontaneous STIM1 activation.Whereas SARAF N-terminus dictates its activity, its C-terminus interacts directly with STIM1.Furthermore, SARAF facilitates the return of STIM1 to the resting state.This is associated with slow Ca 2+ -dependent inactivation (SCDI) (Jha et al., 2013;Maleth et al., 2014;Palty et al., 2012).SARAF binds STIM1 at the ID downstream of the SOAR domain.The STIM1-SARAF co-regulation is achieved through a complex interplay of SARAF with distinct regions of the CTID and is sensitive to PIP 2 .Impaired interplay of STIM1 with PIP 2 leads to a reduction of SARAF-mediated SCDI of STIM1-Orai1 currents.The co-regulation of STIM1 and SARAF is further facilitated by the extended synaptotagmin 1 (ESyt1) and Septin4, which retain STIM1 and Orai1 in PIP 2 -rich regions (Maleth et al., 2014).It is assumed that the STIM1-Orai1 complex can be modulated by transient shuttling between PIP 2 -depleted and -rich regions.SARAF only interacts with STIM1 at high PIP 2 levels to trigger efficient STIM1-Orai1 signalling, as visible by SCDI and reduction of Ca 2+ currents (Maleth et al., 2014), which is in accord with the PIP 2 -mediated modulation of STIM1 function (Calloway et al., 2011).

STIMATE/TMEM110. STIMATE
(encoded by TMEM110) is a multi-TM protein localized in the ER with a polybasic region at the end of the C-terminus.It physically interacts with STIM1-CC1 to facilitate the conformational change of STIM1 to the active state (Jing et al., 2015).Thus, it acts as a positive regulator of STIM1-mediated Orai1 activation.It stabilizes ER-PM junctions required for the interplay between STIM and Orai.It would be interesting to resolve structural changes induced by STIMATE, in particular along the CC1α1-CC3 interface.
J Physiol 602.8

Indirect modulation at the ER-PM junctions
Extended synaptotagmins.Extended synaptotagmins (E-Syt1/2/3), which are situated in the ER and form contacts with the PM via PIP 2 (Chang et al., 2013;Giordano et al., 2013;Kang et al., 2019;Maleth et al., 2014), can modulate STIM1 function depending on the cell type, despite being dispensable for STIM1/Orai1 activation (Giordano et al., 2013).In HeLa cells, E-Syts, in particular E-Syt1, are required to form ER-PM contact sites, which, however, are located at distinct sites compared to those where STIM1 and Orai1 interact.This guarantees efficient Ca 2+ repletion in the ER to bring the cell back into the resting state (Giordano et al., 2013;Kang et al., 2019).In contrast to HeLa cells, in T cells, in particular the expression of a short isoform of E-Syt2, E-Syt2S, which functions in co-regulation with E-Syt1, positively modulates STIM1 clustering and CRAC channel function.E-Syt2S recruits STIM1 in a PIP 2 -dependent manner by direct interaction to ER-PM junctions.A long E-Syt2 isoform, E-Syt2L, is suggested to inhibit STIM1 recruitment to the PM (Woo et al., 2020).
E-Syt1 further supports the recruitment of the lipid transfer protein Nir2 (PYK2 N-terminal domain-interacting receptor 2; Chang et al., 2013).It triggers the exchange of phosphoinositides and phosphatic acids at ER-PM contact sites (Balla, 2018;Kim et al., 2015;Yadav et al., 2015) and maintains PM-PIP 2 levels (Balla, 2018;Chang & Liou, 2015;Chang et al., 2013;Kim et al., 2015;Yadav et al., 2015).Store-dependent translocation of E-Syt1 to the ER-PM junctions facilitates the formation of new ones.This enhances the translocation of Nir2 to and the accumulation of PIP 2 at the PM (Chang & Liou, 2016;Chang et al., 2017;Kim et al., 2015).Elevated PIP 2 levels likely facilitate the colocalization of E-Syt1, Nir2 and STIM1 at the ER-PM contact site to prepare the cell for efficient activation (Chen et al., 2017;Dickson, 2017).
GRAMD2A.GRAM proteins, specifically GRAMD1A and GRAMD2A, represent another type of protein located at ER-PM junctions to facilitate lipid transport (Begley et al., 2003;Berger et al., 2003;Gatta et al., 2015;Murley et al., 2015;Naito et al., 2019).Recently, these proteins were identified to colocalize with STIM1 at ER-PM contact sites in a PIP 2 -dependent manner.In particular, GRAMD2A impacts the formation of STIM1 puncta, while they are not required to induce store-operated Ca 2+ entry (Besprozvannaya et al., 2018).While their detailed functional role still requires further investigations, GRAMD2A proteins are suitable markers for ER-PM contact sites.
ANO8.Another interesting ER-PM tether at PIP 2 -rich regions is the anoctamin ANO8, predicted to be composed of 10 TM domains situated in the ER membrane and a long C-terminus pointing into the cytosol with a series of basic residues assumed to form a PIP 2 binding motif.ANO8 controls STIM1-STIM1 clustering, STIM1-mediated Orai1 activation and inactivation dependent on PIP 2 .It was assumed that ANO8 assembles all essential CRAC channel proteins and associated accessory proteins at the PIP 2 -rich sites at ER-PM junctions to establish efficient Ca 2+ signalling (Jha et al., 2019).More detailed investigations are still required to understand the underlying molecular mechanisms for ANO8 action on the CRAC channel components.
RASSF4.PIP 2 levels, affecting STIM1 function, are further controlled by RASSF4, a RAS association domain (Chan et al., 2013).RASSF4 directly interacts with the adenosine diphosphate ribosylation factor 6 (ARF6), a small G protein, which modulates the generation of phosphatidylinositol-4-phosphate 5-kinase and PIP 2 .In this way it controls STIM1 function in an indirect manner via modulation of PIP 2 levels (Chen et al., 2017).

Future perspectives
Overall, structural resolution, biochemical, biophysical and cellular observations together with computer simulation approaches have attempted to unravel the complex STIM activation machinery.The fascinating and enormous amount of work on various fragments and potential intermediate states of STIM has provided a comprehensive picture of the activation mechanism.However, for some structural resolutions, it is still not known whether they represent an intermediate state and how they fit into the complex restructuring of STIM1 from the quiescent to the active state.Furthermore, discrepancies in the precise location of the apex relative to the ER membrane, and thus the coiled-coil domain interfaces, remain to be clarified.Also, the luminal oligomerization domain as well as the STIM-Orai binding interface still leave many questions unanswered.In particular, much effort is still needed to fully elucidate the pathway of structural reorganization of STIM1 from the resting state to the active state.Thus, on the one hand, a detailed picture of the full-length Ca 2+ -bound as well as Ca 2+ -depleted active protein is still lacking.On the other hand, the structural resolution of STIM in complex with Orai is still pending to provide new insights into the final active state of STIM1, the STIM1-Orai1 binding pockets, and the stoichiometry within the CRAC channel complex.The fact that the number of STIM splice variants is steadily increasing suggests that this allows cell type-specific control of store-operated Ca 2+ entry.However, the molecular basis for the different functional roles of the various isoforms still requires intensive structural and functional investigation.Moreover, accessory proteins and post-translational modifications increase the complexity of this highly controlled STIM-Orai system and enable specific targeting to certain cellular signalling pathways.The elucidation of the basic activation mechanisms of STIM proteins is on the one hand essential to understand physiological processes, and on the other hand of immense importance in the understanding of disease development.A number of disease-related mutations were already characterized, but there are still many unanswered questions, which may play an important role in the development of new therapeutic approaches.

0
Matthias Sallinger is a PhD student in the Ion Channel group at the Institute of Biophysics, Johannes Kepler University (JKU) Linz under the co-supervision of Professor Christoph Romanin and Professor Rainer Schindl.He became interested in biophysical studies and protein science during his undergraduate and master's degrees in Molecular Biology.His research aims to study mechanisms of STIM and Orai protein activation and downstream signalling events in human cells.Herwig Grabmayr recently finished his PhD in the Ion Channel group under the supervision of Professor Christoph Romanin.He has since continued his mechanistic investigations of STIM proteins as a postdoctoral researcher.Rainer Schindl is a biophysicist and Associate Professor at the Medical University of Graz.He is an expert in Ca 2+

Figure 1 .
Figure 1.STIM activation cascade Schematic representation of the STIM activation cascade together with all currently available structural resolutions of STIM as derived from the Protein Data Bank (PDB) with the indicated identifiers.The structures are matched with different stages of STIM activation (left: resting conditions; right: store-depleted, activated conditions).

Figure 2 .
Figure 2. Structural resolutions of STIM N-terminus Currently available NMR solution structures of STIM N-terminal fragments, including (A) human STIM1 (PDB: 2K60), (B) human STIM2 (PDB: 2L5Y), and (C) C. elegans STIM (PDB: 6PW7).Differences in the orientation of α2 and α10 as well as the amino acids comprising the hydrophobic pocket of the specific fragments are depicted.

Figure 3 .
Figure 3. Interactions within STIM1 C-terminus and modelling of STIM1 A, interhelical dimerization sites of the SOAR crystal structure (PDB: 3TEQ).B, interhelical interaction sites of CC3 and CC1α1.C, interhelical interaction sites of CC1α3 and CC1α2.The residues of the individual monomer domains are highlighted in red and green, respectively.D, NMR solution structure of CC1 (PDB: 6YEL).Identified interhelical interactions are indicated in red (CC1α1), blue (CC1α2) and green (CC1α3).E, amino acids potentially involved in inter-and intrahelical interactions of the CC2, apex and CC3 regions within SOAR are shown and highlighted in red.F, in silico model of STIM1 (aa 63−443).This model includes the NMR resolved luminal domain of STIM1 (PDB: 2K60; Stathopulos et al., 2008) as well as the SOAR crystal structure (PDB: 3TEQ; Yang et al., 2012) and considers the configuration determined by van Dorp et al. (2021).The whole STIM1 protein was reconstructed using Modeller software(Fiser & Sali, 2003), which allowed integration of the different subdomains and connecting of the luminal and cytosolic domains with a modelled TM domain.The TM domain and CC1α1 were modelled as a continuous α-helix, up to residue 271.This new subdomain was then apposed to CC3 from SOAR to maximize the contact surface between hydrophobic residues.CC1α2 and CC1α3 were also modelled as α-helices and positioned on top of the CC1α1/SOAR construct.The short sequences of amino acids in between were considered unstructured and modelled as random coils.The CC1α3 domain was then connected to the SOAR crystal structure by adding residues 337-345 in a configuration corresponding to an α-helix.To generate a dimer, SOAR of the first STIM1 monomer was used to position SOAR of the second STIM1 monomer.Specifically, the dimer was created by aligning the SOAR Cα atoms from the first fully reconstructed monomer to the SOAR Cα atoms of the second monomer.A short MD simulation was then performed where all heavy atoms from the backbone remained fixed to relax the side chains.Subsequently, a 1 ns-long molecular dynamics simulation without any constraints was performed to relax the structure.G, graphical representation of residues identified as being important for STIM dimerization (from TM domain in violet to CC1α1, CC1α2 and CC1α3 in blue).

Figure 4 .
Figure 4. Disease-related mutations in STIM1Currently identified and published disease-related STIM1 mutations within the N-terminal domains (A) and C-terminal domains (B).GoF mutations are highlighted in red, LoF mutations are highlighted in blue and mutations whose function is currently unknown are highlighted in violet.
indicate that the non-canonical EF-hand domain is more sensitive to mutation-based