Evidence for an interaction between Golli and STIM1 in store-operated calcium entry

Ca²⁺ signalling controls many cellular processes including cell growth, secretion and proliferation [1,2]. Generation of Ca²⁺ signals in non-excitable cells involves rapid IP₃ (inositol 1,4,5-trisphosphate)-dependent Ca²⁺ release from intracellular stores followed by external Ca²⁺ entry through SOCCs (store-operated Ca²⁺ channels) [3–5]. These coupled processes are regulated by two dynamic transmembrane proteins, the ER (endoplasmic reticulum) Ca²⁺ sensor protein STIM1 (stromal interaction molecule 1) [6–10] and the PM (plasma membrane) protein Orai1 [9,11]. STIM1 is a Ca²⁺ sensor within the ER that oligomerizes to form aggregates or ‘puncta’ in ER–PM junctions upon store depletion [7,12]. Oligomerized STIM1 then binds to PM Orai1 to activate Ca²⁺ influx [13,14]. Mutational analysis of the STIM1-interacting domains revealed that the cytoplasmic C-terminus of STIM1 is necessary and sufficient for the activation of SOCC influx and its associated Ca²⁺ release-activated Ca²⁺ current (I_CRAC) [15]. More recently, a minimal region has been identified within the C-terminus of STIM1, which is required to interact with and activate Orai1 clustering and SOCC influx [16–20].

A previous paper by Varnai et al. [21] suggested that STIM1 and Orai1 form part of a large macromolecular complex, which may indicate the presence of other proteins within STIM1–Orai1 clusters. One potential candidate for this is Golli protein, encoded by the MBP (myelin basic protein) gene; Golli proteins are alternative splice isoforms of the classic MBP proteins [22]. Unlike MBP, which is a major constituent of myelin in the nervous system, Golli proteins are expressed ubiquitously, although their function is not well understood. The BG21 isoform of Golli (Golli-BG21) is expressed at high levels in T-cells [23] where it acts as a negative regulator of T-cell activation [22]. T-cells from Golli-deficient mice are hyperproliferative and exhibit increased IL (interleukin)-2 production [24]. It has been shown that this is due to enhanced SOCC influx in Golli-deficient cells, suggesting that Golli is a negative regulator of SOCE (store-operated Ca²⁺ entry) [24]. Conversely, overexpression of Golli-BG21 in Jurkat T cells results in decreased SOCC influx [24]. Golli-BG21 also regulates voltage-gated Ca²⁺ channels and SOCCs to mediate the migration and proliferation of oligodendrocyte precursor cells [25–27]. Interestingly, a fraction of Golli-BG21 is constitutively localized to lipid rafts at the PM and this PM association increases following T-cell stimulation [24]. Mutation of the N-terminal Golli myristoylation site, required for PM association, abolishes its ability to regulate Ca²⁺ influx [24]. Golli-BG21 does not affect the IP₃-dependent release of Ca²⁺ from internal stores, but rather regulates store-depletion-induced Ca²⁺ entry itself [24]. It is not yet known how Golli-BG21 regulates SOCC activity in T-cells, i.e. whether it is due to a direct effect on the channel itself or through the regulation of the signalling pathway responsible for SOCC activation. In the present study we show that Golli-BG21 interacts with the SOCC activator STIM1 and suggest that down-regulation of SOCE is the likely consequence of this interaction.

All expression vectors were purchased from Clontech, except pGEX-6-P1, which was from GE Healthcare. The Golli-BG21–GFP (green fluorescent protein) plasmid was a gift from Dr Celia Campagnoni (University of California, Los Angeles, School of Medicine, Los Angeles, CA, U.S.A.). For affinity chromatography and binding experiments, the C-terminal domain of STIM1 (STIM1-CT; amino acids Met²⁴¹– Lys⁶⁸⁵), the C- and N-terminal domains of Orai1 [Orai-CT (amino acids Ala²⁶⁷–Ala³¹⁵) and Orai-NT (amino acids Met¹–Leu⁸⁶)] and the full-length Golli-BG21 protein were subcloned into the pGEX-6-P1 vector using BamHI/SacII for STIM1-CT, and BamHI/XhoI for Golli and Orai1. For split YFP (yellow fluorescent protein) experiments, STIM1-CT and Golli-BG21 were subcloned into YFP-YN (N-terminal half of YFP) and YFP-YC (C-terminal half of YFP) vectors respectively [28]. STIM1-CT was subcloned into pEYFP-N1 and YFP-YN using HindIII/SacII and Golli-BG21 was subcloned into pmCherry-N1 (a gift from Dr Roger Y. Tsien, Department of Pharmacology, University of California, San Diego, CA, U.S.A.) and YFP-YC using XhoI/SacII for co-localization experiments. Construction of STIM1–EYFP (enhanced YFP) and mCherry–Orai1 has been described previously [29], as has construction of pmCherry–Rit1 [30].

Extracts of bovine brain were prepared as described previously [31]. GST (glutathione transferase) fusion proteins were expressed in Escherichia coli, immobilized on to glutathione–cellulose beads (Bioline) and incubated with 1 ml of extracted cytosolic bovine brain proteins on a rotator for 2 h at 4 °C. The beads were washed with wash buffer [25 mM Tris/HCl, pH 7.8, 50 mM KCl and 1 mM DTT (dithiothreitol)] and proteins eluted by boiling in Laemmli buffer. Eluted proteins were purified by methanol precipitation, denatured by boiling in SDS buffer, and analysed by SDS/PAGE. Protein bands of interest were excised and subjected to in-gel tryptic digestion and MALDI–TOF MS as described previously [32].

GST control or GST–Golli fusion proteins were immobilized on to glutathione–cellulose beads and incubated with 2 μM STIM1-CT protein, which had been treated with the PreScission protease (GE Healthcare) to cleave the GST tag. Samples were eluted by boiling in Laemmli buffer and analysed by Western blotting using an anti-STIM1 antibody (1:1000 dilution; ProSci) and a HRP (horseradish peroxidase)-linked anti-(rabbit Ig) secondary antibody.

HeLa cells were maintained as described previously [29]. Cells were transfected with 1 μg of the plasmid(s) of interest using GeneJuice (Merck Biosciences) and cultured for 24 h.

For BiFC assays, HeLa cells were transfected with Golli-BG21–YC and STIM1-CT–YN constructs either alone or together. For further control experiments, HeLa cells were transfected with the empty BiFC vectors (YFP-YC and YFP-YN), Golli-BG21–YC and YFP-YN, or STIM1-CT–YN and YFP-YC. Cells were treated with 2 μM thapsigargin at room temperature (23 °C) for 90 min before imaging. For co-localization studies, HeLa cells were co-transfected with mCherry–Orai1 together with Golli-BG21–YC/STIM1-CT–YN or STIM1-CT–YFP, STIM1–EYFP and mCherry–Golli, or triply transfected with STIM1–EYFP, mCherry–Orai1 and Golli-BG21–GFP and imaged 24 h post-transfection.

Cells grown on glass-bottomed dishes were placed in standard buffer (10 mM Hepes, pH 7.3, containing 140 mM NaCl 4.7 mM KCl, 2 mM CaCl₂, 1.13 mM MgCl₂ and 10 mM glucose) and loaded with 5 μM Fluo 4 AM (Fluo 4 acetoxymethyl ester) for 30 min at room temperature before washing in Ca²⁺-free standard buffer. Ca²⁺ imaging was carried out using a Leica AOBS SP2 microscope. HeLa cells expressing STIM1–EYFP, Golli–mCherry or a combination of both were pretreated with 2 μM thapsigargin (Calbiochem) in the absence of extracellular Ca²⁺ for 10 min to deplete stores; 2 mM Ca²⁺ was then added to activate SOCE [30].

Quantification of STIM1–EYFP puncta was performed as described previously [30]. Briefly, puncta were selected as spots of high fluorescence intensity ranging from approx. 0.5–1.0 μm in diameter. Accuracy of puncta quantification was verified by independent blind counting.

To determine the extent of Golli–mCherry fluorescence in cells overexpressing STIM1–EYFP compared with cells that do not overexpress STIM1, regions of interest were drawn around the outside of each cell. Fluorescence values were calculated on the basis of the mean fluorescence per pixel within the cell rather than total fluorescence to avoid problems arising from variations in the size of the selected regions of interest.

To search for novel binding partners for STIM1 and Orai1, we performed pull-down experiments on immobilized cytosolic domains of the STIM1 and Orai1 GST-fusion proteins [GST–STIM-CT (~80 kDa), GST–Orai1-CT (~34.5 kDa) and GST–Orai1-NT (~33 kDa)] using bovine brain cytosol (Figure 1A). MALDI–TOF MS analysis of unique protein bands from the pull-down identified MBP (eight matching peptides with 39% sequence coverage), as a protein of approx. 25 kDa in size, which bound to the C-terminus of STIM1. Western blotting using an anti-MBP antibody confirmed that MBP bound specifically and efficiently to STIM1 and also bound to the N-terminus of Orai1 (Figure 1B). Note that a similar sized polypeptide was present in the control GST pull-down sample (Figure 1A). The identity of this polypeptide was not established, but it was not recognized by the anti-MBP antibody (Figure 1B). The finding of MBP binding was surprising, since classic MBP itself is required for central nervous system myelination and has not been implicated in the activity of SOCCs. The Golli family of proteins are alternatively spliced variants of classical MBPs, which have been shown to mediate signal transduction in T-cells through the negative regulation of SOCCs [24]. As Golli proteins contain MBP epitopes, we hypothesized that Golli may interact with STIM1 to exert its effect on SOCE. We assessed whether the C-terminus of STIM1 interacts directly with Golli using an in vitro binding assay. A recombinant GST–Golli-BG21 fusion protein (~57.5 kDa) was expressed in E. coli and immobilized on to a glutathione affinity resin. Immobilized GST–Golli-BG21 or GST control protein (~26 kDa) was incubated with PreScission-cleaved STIM1CT (~54 kDa). Western blotting analysis revealed that STIM1-CT bound specifically to the GST–Golli-BG21 protein, but not to the GST control (Figure 1C), confirming that the C-terminus of STIM1 does indeed interact with Golli in vitro. Note that in these experiments the glutathione beads were incubated with levels of GST or GST–Golli-BG21 that resulted in a 1.4-fold higher level of GST so that lack of binding of STIM1-CT to GST-loaded beads was not a consequence of the relative GST concentration. In similar experiments binding of Orai-NT to GST–Golli-BG21 could not be detected (results not shown).

To investigate further the possible interaction of STIM1 and Golli-BG21 in a cellular context, we expressed these two proteins in HeLa cells using a BiFC assay [33]. This BiFC system involved the fusion of the complementary fragments of EYFP (i.e. YN and YC) to two proteins of interest. If these proteins interact the EYFP fragments come close enough to fold and form a functional fluorescent EYFP protein [28]. In these experiments, YC and YN were fused to the C-termini of the STIM1-CT (STIM1-CT–YN) and Golli (Golli-BG21–YC) proteins and expressed in HeLa cells along with a mCherry-tagged Rit1 (Ras-like without CAAX 1) tail protein (mCherry–Rit1), described previously [30], as a marker to identify transfected cells. No detectable fluorescence was observed in untreated cells (Figure 2A). Transfected cells were treated with thapsigargin for 60–90 min at room temperature in 2 mM external Ca²⁺ before imaging with a confocal microscope. EYFP fluorescence was detected at the PM in treated cells expressing STIM1-CT–YN and Golli-BG21–YC at approx. 90 min following the addition of thapsigargin (Figure 2B), which is consistent with the time course described for the refolding of EYFP at room temperature [33]. EYFP fluorescence was distributed throughout the PM, but no puncta were visible. The uniform PM EYFP fluorescence observed was consistent with the previous finding that although STIM1-CT expression gives constitutive activation of the I_CRAC, it does not form puncta at the PM [34] as it requires an ER localization to oligomerize prior to translocation to ER–PM junctions [34,35]. We also observed that STIM1-CT–EYFP co-localizes with the mCherry–Orai1 protein at the PM in HeLa cells, but does not form visible puncta (results not shown). No detectable EYFP fluorescence was observed in thapsigargin-treated cells expressing either the STIM1–YN or Golli-BG21–YC fusion protein alone (Figures 2C and 2D) or along with the appropriate complementary fragment of YFP (Figures 2E and 2F). Similarly, co-expression of both EYFP fragments failed to produce a fluorescent signal and served as a negative control (Figure 2G). There are two possibilities to explain why EYFP fluorescence was observed at the PM only following thapsigargin treatment. First, STIM1-CT–YN may associate with endogenous STIM1 puncta enabling it to interact with Golli-BG21–YC at the PM. Secondly, it is also possible that a change in cytosolic Ca²⁺ concentration may be required to modulate an interaction between STIM1-CT–YN and Golli-BG21–YC. To investigate this, HeLa cells expressing both constructs were treated with 100 μM histamine, which should induce transient oscillations in cytoplasmic Ca²⁺, in place of thapsigargin. Plasmalemmal EYFP fluorescence formation was observed (Figure 2H) even though full-length STIM1–EYFP did not reveal substantial sustained puncta formation under these conditions (results not shown) suggesting that an increase in cytosolic Ca²⁺ from intracellular stores modulates the interaction between STIM1-CT and Golli-BG21. PM EYFP fluorescence was also observed in cells that were not co-transfected with mCherry–Rit1 (results not shown), eliminating the possibility that the transfection marker was targeting the BiFC constructs to the PM. The emission spectrum for the fluorescence observed with STIM1–YN/Golli-BG21–YC was close to that of EYFP (Figure 2I), confirming that the fluorescence observed was due to formation of functional EYFP. The BiFC assay was also used to determine whether Golli interacts with full-length STIM1–YN or Orai1–YN. However, EYFP fluorescence was not observed in cells transfected with Golli–YC and Orai1–YN or full-length STIM1–YN that had been treated with thapsigargin for 90 min (results not shown).

Interestingly, when mCherry-Orai1 was co-expressed with STIM1-CT–YN/Golli-BG21–YC in HeLa cells, the EYFP fluorescence clearly overlapped with the mCherry–Orai1 signal (Figure 3A). Furthermore, a Golli-BG21–GFP fusion co-localized with both mCherry–Orai1 and STIM1-CT–EYFP in HeLa cells (Figure 3B and insert). This led us to examine whether Golli co-localizes with full-length STIM1 and Orai1 in puncta, thereby placing the protein in a position to regulate the activity of these proteins. Golli overexpression has been shown to decrease SOCE following store depletion in Jurkat T cells [24]. If this occurred through an interaction with STIM1, the Golli protein would either co-localize with STIM1 in puncta or would prevent STIM1 puncta formation. To test this, full-length STIM1–EYFP and a Golli-BG21–mCherry fusion protein were expressed in HeLa cells. No co-localization between STIM1–EYFP and Golli-BG21–mCherry was observed in untreated HeLa cells, where STIM1–EYFP expression was reticular and Golli-BG21–mCherry was distributed throughout the PM as observed previously [24] (Figure 3C). However, when cells were treated with thapsigargin for 10 min before imaging to induce STIM1 puncta formation, Golli–mCherry expression was observed predominantly at the PM where STIM1–EYFP puncta accumulated, but was also seen to co-localize with STIM1–EYFP in a subset of puncta (Figure 3D and insert). We also examined whether Golli–GFP co-localized with punctate STIM1–EYFP and mCherry–Orai1 complexes in HeLa cells after thapsigargin treatment. Indeed it appeared that all three proteins co-localized to the same punctate structures (Figure 3E and insert). We did not see any change in the number of STIM1–EYFP puncta formed in thapsigargin-treated cells overexpressing Golli–mCherry compared with control cells expressing STIM1–EYFP alone (Figure 3F) suggesting that Golli does not affect the formation of STIM1 puncta itself. The targeting of both proteins to the same punctate structures suggests that Golli and full-length STIM1 interact and that Golli may regulate SOCE through a direct interaction with STIM1 in puncta.

To determine whether Golli affects SOCE in HeLa cells, we transfected cells with Golli–mCherry or STIM1–EYFP alone or in combination and loaded cells with the cytosolic Ca²⁺ indicator, Fluo 4 AM. Stores were depleted with thapsigargin in Ca²⁺-free medium for approx. 10 min and Ca²⁺ influx was measured upon re-addition of Ca²⁺ to the external solution. Untransfected cells in the same field of view as transfected cells were used as internal controls. Golli-expressing cells exhibited a reduced rate and extent of Ca²⁺ influx compared with influx in untransfected control cells (Figure 4A), which was statistically significant (P<0.0001) at the 750 s time point. This is consistent with previous findings that Golli-deficient T-cells show enhanced SOCE and that overexpression of Golli-BG21 inhibits SOCE [24], and suggests that Golli negatively regulates SOCE. In STIM1-overexpressing cells there was an increase in the initial rate of SOCE when compared with untransfected cells (Figure 4A), which we have observed previously [30]. Interestingly, overexpression of STIM1 abolished the inhibitory effect of Golli-BG21 when compared with STIM1-overexpressing cells (Figure 4A). This was not due to differences in the expression of Golli as the level of Golli-BG21–mCherry fluorescence was similar in cells expressing Golli-BG21–mCherry alone or in combination with STIM1–EYFP (Figure 4B). It is therefore possible that high levels of STIM1 protein expression can directly overcome the inhibition of SOCE by Golli-BG21.

Previous studies identified Golli-BG21 as a regulator of SOCE. This has been particularly well established for T-cells where knockout of Golli increased SOCE and overexpression of Golli-BG21 decreased SOCE [24]. It was not clear, however, from these studies whether Golli has a direct effect on the SOCC itself or another protein or pathway involved in the activation of SOCE. In the present study, we suggest that Golli directly interacts with the master SOCE regulator STIM1 to reduce SOCE. The C-terminal domain of STIM1 successfully bound to Golli in GST-binding assays and BiFC assays. We have not mapped the STIM1-interacting domain within Golli, but it is possible that this is an epitope that is shared with the classical MBP protein given the C-terminus of STIM1 binds to MBP from bovine brain extract in pull-down assays. We have shown in the present study, using a BiFC assay, that the interaction between Golli-BG21 and the C-terminus of STIM1 can be detected in cells after stimulation by both thapsigargin and histamine treatment. Both of these compounds should produce changes in cytosolic Ca²⁺ levels but, unlike thapsigargin, histamine elicits transient Ca²⁺ oscillations in HeLa cells [36,37]. As the formation of STIM1–EYFP puncta was not observed under conditions where histamine stimulated BiFC fluorescence between STIM1–YN and Golli–YC, it is possible that it is the modulation of the cytosolic Ca²⁺ concentration by both thapsigargin and histamine rather than sustained translocation of endogenous STIM1 that induces an interaction between the C-terminus of STIM1 and Golli-BG21. The BiFC assay was used in the present study since it is a more robust method for investigating protein interactions in live cells than other techniques such as FRET (fluorescence resonance energy transfer). It is important to note, however, that the BiFC assay is not reflective of the kinetics of the interaction, since binding between the YC and YN fragments of EYFP is essentially irreversible and refolding of the full EYFP protein and the reconstitution of fluorescence requires some time [33], in this case, up to 90 min. Irrespective of the lack of information from these experiments on the kinetics of the STIM1–Golli-BG21 interaction, this assay provides the most convincing evidence of a direct protein–protein interaction in living cells.

No fluorescence complementation was observed in cells expressing Golli–YC and Orai1–YN, or Golli–YC and full-length STIM1–YN (results not shown). EYFP fluorescence was also absent in cells expressing Orai1–YC and STIM1-CT–YN (results not shown), even though the STIM1 and Orai1 proteins are known to interact [16,34,38]. There are several reasons why this may have occurred. First, fusion of an EYFP fragment to a protein may alter the structure of the protein, thereby abolishing an interaction domain within the protein of interest [39]. Secondly, the arrangement of the BiFC fusion proteins within a protein complex may sterically hinder an interaction between the complementary EYFP fragments and prevent the formation of functional EYFP [39]. The lack of EYFP fluorescence between Golli–YC and Orai1–YN, and Golli–YC and full-length STIM1-YN proteins is therefore not conclusive evidence for the absence of an interaction between these protein pairs. We also could not confirm the potential interaction of Orai-NT to GST–Golli-BG21, seen in the initial GST pull-down from brain extract, in a direct binding assay with recombinant proteins.

Despite the large amount of recent research into the activation of SOCCs, very few interacting partners have been found for STIM1 and Orai1. One recent study has identified a novel EF-hand-containing protein CRACR2A (CRAC regulator 2A), which binds to both STIM1 and Orai1 to stabilize their interaction, thereby enhancing SOCE in T-cells [40]. Additionally, little is known about the Ca²⁺-dependent inactivation of these channels. One recent study reported that both calmodulin and a short negatively charged region within the STIM1 C-terminus can bind to Orai1 to inactivate SOCE [41]. Interestingly, in the present study, Golli-BG21 co-localized with STIM1 and Orai1 complexes after store depletion. The ability of Golli-BG21 to bind STIM1 puncta in a Ca²⁺-dependent manner may be key to its regulation of SOCE. It is possible that Golli-BG21 binds to STIM1–Orai1 complexes when Ca²⁺ levels are already high to reduce the influx of Ca²⁺ into the cell and perhaps may be a further contributor to the Ca²⁺-dependent inactivation of SOCCs. When STIM1 was overexpressed with Golli-BG21, the inhibitory effect of Golli-BG21 on SOCE was abolished. This was not due to variation in the expression levels of Golli-BG21 achieved in the presence and absence of STIM1, suggesting that an interaction between STIM1 and Golli-BG21 overcomes the ability of Golli to reduce SOCE.

To date, studies on MBP and Golli, and their regulation of Ca²⁺ channels, have focussed exclusively on oligodendrocyte precursor cells and T-cells. Of the three Golli isoforms cloned, the BG21 isoform of Golli is the most widely expressed in many different tissues, other than nervous tissues, including the heart, kidney, spleen and lung [23], although its function in these tissues has not been elucidated. The widespread distribution of Golli-BG21 suggests that it could be part of a general mechanism for the regulation of SOCE across many tissue types. Hence, we propose that Golli-BG21 functions to regulate SOCE via a direct interaction with STIM1, but the exact molecular mechanism underlying this interaction has yet to be defined.

BiFC

bimolecular fluorescence complementation

CT

C-terminal domain

ER

endoplasmic reticulum

Fluo

4 AM, Fluo 4 acetoxymethyl ester

GFP

green fluorescent protein

GST

glutathione transferase

I_CRAC

Ca²⁺ release-activated Ca²⁺ current

IP₃

inositol 1,4,5-trisphosphate

MALDI–TOF

matrix-assisted laser-desorption ionization–time-of-flight

MBP

myelin basic protein

NT

N-terminal domain

PM

plasma membrane

Rit1

Ras-like without CAAX 1

SOCC

store-operated Ca²⁺ channel

SOCE

store-operated Ca²⁺ entry

STIM1

stromal interaction molecule 1

YFP

yellow fluorescent protein

EYFP

enhanced YFP

YC

C-terminal half of YFP

YN

N-terminal half of YFP

Ciara Walsh contributed to the experimental design, carried out the experimental work and prepared the first draft of the paper. Mary Doherty performed the MS and related experiments. Alexei Tepikin contributed to the conception of the project, to the experimental design and edited the paper. Robert Burgoyne contributed to the conception of the project, to the experimental design and the preparation of the final version of the paper.

We thank Dr Celia Campagnoni (deceased 2010) and Dr Roger Y. Tsien for the gifts of plasmids used in the present work.

This work was supported by a Wellcome Trust Prize Ph.D. studentship [grant number 080910/Z/06/Z (to C.M.W.)]