Agonist-activated Ca²⁺ influx occurs at stable plasma membrane and endoplasmic reticulum junctions

In excitable and non-excitable cells, activation of Ca²⁺ release from stores present in the endoplasmic or sarcoplasmic reticulum is accompanied by Ca²⁺ influx from the extracellular environment, a fundamental step that is necessary to replenish intracellular stores and allow cells to respond to successive stimulatory signals (Feske et al., 2001; Putney, 2007; Berridge, 2004; Parekh and Putney, 2005; Eisner and Trafford, 2009). In non-excitable cells, the molecular mechanism involved in store-operated Ca²⁺ entry (SOCE) has been extensively investigated and a number of molecules have been shown to be involved in Ca²⁺ influx (Wu et al., 2007; Putney, 2007). The most popular candidates expressed in most mammalian cells are STIM1 (stromal interaction molecule 1) and Orai1 (calcium-release-activated calcium channel protein 1). STIMs are proteins that contain a single transmembrane domain and a luminal N-terminal domain endowed with an EF hand that functions as a Ca²⁺ sensor in the lumen of the endoplasmic reticulum (ER) (Liou et al., 2005; Zhang et al., 2005; Stathopulos et al., 2006; Spassova et al., 2006; Huang et al., 2006); Orai1 is the pore-forming unit of the SOCE channel (Gwack et al., 2007; Prakriya et al., 2006; Yeromin et al., 2006). When intracellular Ca²⁺ stores are full, STIM1 proteins are diffusely distributed in the ER as monomers, but upon store depletion, STIM1 undergoes conformational changes resulting in oligomerization and redistribution into puncta located 10–25 nm from the plasma membrane (PM) (Luik et al., 2008; Vig et al., 2006; Wu et al., 2006; Park et al., 2009). Orai1 proteins are then recruited to STIM-containing puncta, whereby they form a complex and reconstitute functional Ca²⁺ influx channels (Wu et al., 2007; Park et al., 2009; Penna et al., 2008). The time frame of STIM1 relocalization is in the order of hundreds of seconds (Wu et al., 2006) and appears to be of the same order of magnitude of the peak inward Ca²⁺ current via I_CRAC (Hofer et al., 1998; Wu et al., 2006). However, recent studies have suggested that SOCE is a complex mechanism that might require additional components (Varnai et al., 2007; Gwozdz et al., 2008).

Agonist-mediated generation of inositol 1,4,5-trisphosphate [Ins(1,4,5)P₃] induces inositol 1,4,5-trisphosphate receptor (InsP₃R)-dependent Ca²⁺ release from intracellular stores, and this is followed by sustained Ca²⁺ entry, an event known as agonist-activated Ca²⁺ entry (Berridge, 1995). This is mediated by the influx of extracellular Ca²⁺ via receptor- and/or store-operated Ca²⁺ channels localized on the PM. In principle, agonist-activated Ca²⁺ entry could be a secondary event, mediated by the depletion of InsP₃-sensitive stores, resulting in the activation of SOCE. However, experimental data indicate that agonist-activated Ca²⁺ entry is functionally distinct from SOCE; in fact, Van Rossum and co-workers (Van Rossum et al., 2004) have shown that agonist-activated Ca²⁺ entry occurs in DT40 cells expressing InsP₃R mutants that are unable to release Ca²⁺ from the ER, but are capable of interacting with their ligand, InsP₃, although these data have stirred some controversies (Vazquez et al., 2002; Dellis et al., 2006). If the InsP₃R is the only intracellular Ca²⁺ channel of DT40 cells, these data have two important implications. First, stimulation of agonist-activated Ca²⁺ entry might occur without store depletion and without the translocation of ER membrane domains from the cytosol to the subplasmalemmal space. Second, Ca²⁺ entry is activated by the interaction of InsP₃ with InsP₃Rs and this interaction in turn induces a conformational coupling between InsP₃Rs and Ca²⁺ entry channels, via a mechanism similar to that involved in muscle excitation–contraction coupling. The anatomical site of muscle excitation–contraction coupling is the calcium release unit (CRU) (Franzini-Armstrong and Jorgensen, 1994; Franzini-Armstrong and Protasi, 1997), which contains ordered arrays of ryanodine receptor Ca²⁺ release channels coupled to L-type voltage-dependent Ca²⁺ channels. CRUs might be formed by a stable and direct association of the sarcoplasmic reticulum with the PM (peripheral couplings, mostly found in developing muscles) or by the association of two to three elements, a transverse T tubule, and one or two closely apposed terminal cisternae of the sarcoplasmic reticulum (triads and diads, representing mature CRUs) (Franzini-Armstrong and Protasi, 1997). Interestingly, it was recently shown that in skeletal muscle, Junctophilin-1, a 72 kDa membrane-spanning protein, mediates the physical linkage of transverse tubules to the SR membrane (Takeshima et al., 2000; Weisleder et al., 2008).

Experimental evidence indicates that the structural requirements for SOCE and agonist-activated Ca²⁺ entry are substantially different. In the first case, store depletion induces long-range dynamic interactions between the ER membrane and the PM, leading to the formation of ER–PM junctions; the formation of such transient junctions allows the functional coupling between STIM1 and Orai1 (Park et al., 2009). The agonist-activated Ca²⁺-entry model, however, predicts the existence of pre-formed ER–PM junctions without the requirement of store depletion during receptor-mediated signal transduction (Berridge, 2004). ER–PM junctions might be a macromolecular complex, which contains, in addition to InsP₃R and Ca²⁺ influx channels, scaffold proteins modulating the functional interaction between the InsP₃R and the Ca²⁺-entry channel (Kiselyov et al., 2005). Understanding of the structural organization of the molecular components involved in Ca²⁺ entry is important to define the mechanism(s) by which Ca²⁺ entry is physiologically activated.

To investigate the mechanism(s) involved in modulation of Ca²⁺ entry, we studied the relationship between the structural and functional properties of the supramolecular complex involved in Ca²⁺ entry, which is composed of junctate (one of the splice variants encoded by the ASPH locus), the short transient receptor potential channel TRPC3 and InsP₃R (Treves et al., 2000; Treves et al., 2004). Here, we show that in HEK293 cells stably transfected with junctate–YFP, junctate distributes in the ER as well as in discrete puncta on the plasma membrane, and the latter correlate with the appearance of peripheral couplings between the ER and the PM. The number of ER–PM peripheral couplings is further increased by co-expression of junctate and TRPC3 channels. Interestingly, the gap of the peripheral couplings is regularly occupied by electron-dense structures of 10 nm. TIRF microscopy demonstrates that junctate-containing puncta are involved in agonist-mediated Ca²⁺ influx. Contrary to what has been observed for STIM1, however, activation of Ca²⁺ release from the ER does not result in a rapid movement or redistribution of the junctate-containing ER to the PM. Based upon these results we conclude that agonist-activated Ca²⁺ entry might also occur at stable ER–PM junctions endowed with junctate and other components of the supramolecular complex mediating Ca²⁺ influx, including TRPC3 channels and InsP₃R.

Junctate is a scaffold protein which might be involved in the assembly of peripheral junction between the PM and the ER by forming a supramolecular complex with InsP₃Rs and the Ca²⁺-influx channel TRPC3 (Treves et al., 2000; Treves et al., 2004; Kiselyov et al., 2005) (supplementary material Fig. S1). In the first set of experiments, we examined whether peripheral couplings are endowed with some of the components making up the junctate–InsP₃R–TRPC3 supramolecular complex. For this investigation we obtained stable HEK293 clones expressing the junctate–YFP fusion protein. We screened several clones expressing varying amounts of junctate–YFP and the major functional properties of clone 21 are summarized in Fig. 1. The expression of the junctate–YFP fusion protein was evaluated by western blot analysis using anti-GFP antibodies (Fig. 1A, left). Fig. 1 also shows that the major InsP₃R isoform expressed in our HEK293 cells was type 3 and chronic overexpression of junctate was associated with a significant increase in its level of expression as determined by western blotting and real-time RT-PCR (Fig. 1A). The absence of immunoreactive bands with antibodies against type 1 and 2 InsP₃R in HEK293 microsomes was not due to the lack of reactivity of the antibodies because control western blots on mouse brain microsomes were positive (results not shown). We next monitored global intracellular Ca²⁺ homeostasis by assessing the size of agonist- and thapsigargin-sensitive Ca²⁺ stores of clone 21 in a population of Fura-2-loaded cells. Overexpression of junctate is associated with significantly larger global peak Ca²⁺ transients in response to the addition of ATP, a purinergic receptor agonist that triggers InsP₃ production, or to 400 nM of the SERCA inhibitor thapsigargin (Fig. 1B). Overexpression of junctate did not significantly affect the resting [Ca²⁺], because the mean Fura-2 resting fluorescence ratios of HEK293 cells and clone 21 cells were similar (1.14±0.02 and 1.11±0.02 in 12 and 10 independent measurements, respectively).

We next examined whether the junctate–YFP recombinant fusion protein is localized within a space that is 100–150 nm from the PM. We performed such analysis by TIRF microscopy, a technique that offers higher spatial resolution (150 nm) compared with that offered by conventional confocal microscopy. Fig. 1C shows bright-field images (left panels) and epifluorescence images of HEK junctate–YFP stable clones (top left) or HEK293 cells transiently transfected with a plasmid encoding SScalEGFP (bottom left), a recombinant protein that is targeted to the ER, and is comprised of a signal sequence, followed by GFP and KDEL at its C-terminus. As seen by epifluorescence (Fig. 1C, central panels), the fluorescent patterns of junctate–YFP and SScalEGFP were virtually identical, radiating from the perinuclear area towards the PM in a mesh-like pattern. When the focal plane was fixed to the interface of glass coverslip with the cell membrane using a surface reflection interference contrast filter (SRIC cells, Fig. 1C, right panels) and cells were viewed by TIRF microscopy (488 nm excitation), the fluorescent patterns were significantly different. In fact, junctate–YFP appeared as discrete fluorescent puncta (Fig. 1D, top), whereas the GFP fusion protein targeted to the ER had a worm-like appearance on discrete areas of the PM (Fig. 1D, bottom). The size and shapes of the fluorescent particles were analyzed in greater detail and are summarized in Table 1, the length and radius of the junctate–YFP particles was significantly smaller than those observed in SScalEGFP-transfected cells; furthermore, the shape factor (1.0 is the value of a circle) of the puncta in junctate–YFP-transfected cells was 0.94 and significantly higher than that obtained in SScalEGFP-transfected cells. We also investigated whether the InsP₃R3 in HEK junctate–YFP cells colocalizes in a domain close to the plasma membrane. Detailed analysis by TIRF microscopy (n=21 cells) showed that within the area positive for YFP fluorescence, 19.4±1.3 % was also positive for InsP₃R3, and within the area positive for InsP₃R3, 15.5±1.1% was also positive for GFP (Fig. 1E). These results indicate that a small percentage of junctate and InsP₃R3 colocalize in a domain closely adjacent to the PM.

It is now well established that store-depletion activation of calcium entry (SOCE) depends on the relocalization of STIM1 from the ‘whole’ ER to a domain of the ER in close proximity to the PM. We examined by TIRF microscopy whether ER containing junctate–YFP similarly relocalizes to the PM during agonist stimulation. To avoid artefacts linked to minimal changes of the focal plane during time-lapse recording in the TIRF mode, we equipped our microscope with an on-line focus control. Addition of 100 μM ATP to HEK293 Cl21 cells did not affect the distribution, the appearance or the movement of the junctate–YFP puncta (Fig. 2A, t=0 to t=180 seconds). The area covered by the fluorescent puncta during stimulation with Krebs-Ringer alone or Krebs-Ringer containing 100 μM ATP were not significantly different (0.017±0.0001 and 0.015±0.0001 μm²; 35 puncta and 43 puncta from four and five different experiments, respectively). Similarly, we did not observe relocalization of the fluorescent signal due to SScalEGFP, an artificial ER-resident protein (Fig. 2B) in transfected HEK293 cells after the addition of 100 μM ATP. If junctate–YFP-containing puncta are stably localized within a space which is 100–150 nm from the PM, one would expect that the appearance of stable puncta should be associated with an increase of peripheral junctions between the ER and the PM in cells overexpressing junctate. We tested this idea by performing transmission electron microscope analysis of junctate–YFP-expressing clones. Fig. 2C shows electron micrographs of the cell periphery in HEK junctate–YFP cells and the arrows indicate areas of contact between the PM and the ER. Where the ER membrane was parallel to the plasma membrane, the gap between the two membranes was 10.4±2.9 nm (mean ± s.d., from 13 junctions, 84 measurements). The width was not significantly different from that of HEK293 cells (9.4±3.4 nm from 38 junctions, 229 measurements). A variable portion of the gap (Fig. 2C) was regularly occupied by electron-dense structures of ~20 nm in width and 10 nm in height, which bridge the junctional gap (Fig. 2C, small arrows). Table 2 summarizes the statistical analysis of the number and size of junctions in sectioned profiles of cells. The length of the junctions in HEK293 cells and in the same cells stably expressing junctate–YFP was not significantly different (0.144±0.023 vs 0.159±0.009 μm, respectively); however, the number of junctions per sectioned profile in junctate–YFP-expressing cells was increased by almost threefold (0.092±0.031 vs 0.270±0.053) compared with YFP-expressing clones.

The previous set of data suggests that the appearance of puncta and peripheral junctions is enhanced by imposing high levels of expression of junctate, one of the molecular components of the InsP₃R–TRPC3 supramolecular complex. If this is true, we reasoned that overexpression of more than one of the components of this supramolecular complex might result in a further increase of ER to plasma membrane peripheral junctions. We set out to verify this hypothesis by analyzing the structural properties of HEK cell clones stably overexpressing both TRPC3 and junctate (clone D6) and comparing them with those of clones overexpressing junctate–YFP. Semiquantitative RT-PCR of HEK (Fig. 3A) clone D6 shows that the level of mRNA encoding junctate was higher than that in HEK Cl28 control cells (HEK-T3 cells transfected with the empty pcDNA 3.1/Hygro plasmid); detailed analysis by real-time RT-PCR revealed that the level of expression of junctate was highest in clone 21, followed by clone D6 (Fig. 3C). Electron microscopy images of junctate–TRPC3-overexpressing clone D6 are shown in Fig. 3D and Table 2 summarizes results obtained on the number and size of junctions in all the HEK clones analysed. Similarly to what was observed in stable clones overexpressing junctate–YFP alone, the presence of both proteins did not significantly affect the average length of the PM–ER peripheral junctions, even though some particularly long junctions were observed (Fig. 3D). The junctional gap width (10.0±3.3 nm, from junctate–TRPC3 HEK cells, 211 measurements) was also not significantly different from either HEK or HEK junctate–YFP cells; however, the number of junctions in those cells overexpressing junctate and TRPC3 was four times higher than in those cells overexpressing junctate alone (Table 2). These data strongly suggest that the expression of the components forming the junctate–InsP₃R–TRPC3 supramolecular complex stabilizes the association between the ER and the PM, but does not alter their structure.

Stable ER–PM peripheral junctions might represent anatomical sites that are responsible for entry of extracellular Ca²⁺ into the cytoplasm and thus replenishment of intracellular Ca²⁺ stores occurring after agonist stimulation of InsP₃R. Fig. 3E shows experiments conducted in HEKT3 D6 (junctate +TRPC3 overexpressing cells): addition of 2 mM Ca²⁺ in the absence of an InsP₃-stimulating agonist did not result in Ca²⁺ influx (Fig. 3Ea, small arrow), whereas addition of Ca²⁺ in the presence of ATP (Fig. 3Eb, large arrow) resulted in a robust Ca²⁺ increase. We have previously shown that overexpression of junctate in combination with the TRPC3 entry channel is accompanied by larger intraluminal Ca²⁺ content, which can be mobilized by the addition of either 400 nM thapsigargin or 100 μM ATP (Treves et al., 2004) (supplementary material Fig. S2)

Although HEK293 clones 21 and D6 have a larger number of ER–PM junctions compared with control cells and because we consistently failed to observe Ca²⁺ entry in the absence of the InsP₃R agonist ATP (Fig. 3Ea), it seems that junctate modulates receptor-activated Ca²⁺ entry only in the presence of InsP₃. Indeed, previous reports have demonstrated that agonist-mediated Ca²⁺ entry requires the interaction of the InsP₃R with its ligand (Van Rossum et al., 2004). Since junctate can bind to InsP₃R and TRPC3 to form the junctate–InsP₃R–TRPC3 supramolecular complex, it is plausible that the lack of substantial Ca²⁺ entry in the absence of InsP₃ generation is due either to the interaction of the junctate N-terminus with a domain of the InsP₃R involved in ligand binding, or to the inhibitory interaction of junctate with the TRPC3 channel. To discriminate between these two possibilities we mapped the binding site for junctate on the InsP₃R.

In vitro transcribed and translated overlapping fragments of [³⁵S]Met-labelled human InsP₃R3 (Fig. 4A) were synthesized and incubated with streptavidin-coupled beads coated with a biotinylated peptide encompassing the N-terminus of junctate (Fig. 4B), the N-terminus of aspartyl-β-hydroxylase (Fig. 4C) or an unrelated peptide (Fig. 4D). Only beads that had been coated with the peptide encompassing the N-terminus of junctate interacted with recombinant InsP₃R fragments. In particular, the N-terminus of junctate pulled down construct F1, which encompasses the InsP₃R domain involved in ligand (InsP₃) binding, as well as construct F4, which encompasses residues 1391–2249. A faint band indicating binding to construct F3 (residues 1119–1460) was also visible. The N-terminus of aspartyl-β-OH failed to interact with constructs F1–F5, as did the unrelated peptide.

We next verified whether the N-terminus of junctate bound within the same domain involved in ligand binding, by performing pull-down experiments in the presence of increasing concentrations of InsP₃. For this assay, we tested InsP₃ concentrations that are used to activate the receptor in vitro (Ando et al., 2006). Fig. 5A shows that as the InsP₃ concentration in the binding solution increased from 0.3 μM to 10 μM, the intensity of the F1 band proportionally decreased (top panel); the addition of 1 μM InsP₃ caused a 20% decrease, whereas the addition of 10 μM InsP₃ decreased the binding between F1 and junctate by approximately 60%. Addition of 10 μM ATP (control) had no effect on the interaction between the two polypeptides (Fig. 5A, bottom panel). We further verified the effect of junctate on [³H]InsP₃ binding by performing ligand-binding experiments in the presence of either the peptide encompassing the N-terminus of junctate or an unrelated peptide. Fig. 5B shows that the amount of InsP₃ bound (B_max) at low nanomolar concentrations was similar in the presence of either 50 μM junctate peptide or the unrelated peptide. In summary, these results show that junctate can interact directly with the domain of the InsP₃R that is involved in ligand binding. In the presence of InsP₃, the amount of F1 (residues 1–686 of the human InsP₃R3 isoform) bound to the N-terminus of junctate was significantly reduced. Since ligand binding was unaffected by the presence of the junctate N-terminal peptide, it follows that the affinity of the InsP₃R for its ligand InsP₃ is probably much higher than that for junctate.

The results obtained using TIRF and electron microscopy indicate that junctate–YFP puncta correspond to cell domains endowed with stable peripheral couplings between the ER and the PM, which might form one of the anatomical sites responsible for Ca²⁺ entry via a conformational coupling mechanism. We next measured Ca²⁺ influx by monitoring Fura Red fluorescence changes in TIRF mode in puncta regions. Fura-Red-loaded HEK junctate–YFP cells were first observed by epifluorescence at 488 nm excitation (Fig. 6Ab), with the SRIC filter (Fig. 6Ac) to fix the correct focal plane to be used in TIRF mode, and subsequently the junctate–YFP puncta were visualized by exciting the cells with a Sapphire laser (Fig. 6Ad). Under these conditions the fluorescence resulting from Fura Red was virtually absent because the excitation and emission wavelength spectra of the Ca²⁺ indicator (λex, 410 nm; λem, 620 nm) did not overlap with those of YFP. Keeping the same focal plane fixed for YFP puncta observation, HEK junctate–YFP cells were then illuminated with a laser at 405 nm; at this wavelength, the Ca²⁺-bound form of Fura Red was excited with an efficiency of approximately 70%, whereas there was virtually (<2%) no excitation of YFP. By using emission barrier filters (625 nm), we monitored the emission of Fura Red and thus were able to follow changes in [Ca²⁺]_i in regions within a space of 150 nm of the PM. Fig. 6B shows the changes in the fluorescent ratio caused by the addition of 100 μM ATP in the absence (dotted trace) and in the presence (continuous trace) of 2 mM extracellular Ca²⁺, occurring within a punctuated junctate–YFP-containing area. In the presence of 2 mM extracellular Ca²⁺, the kinetics and magnitude of the Ca²⁺ signal were clearly different from those observed in the presence of EGTA. In the presence of 2 mM Ca²⁺, the signal had a rapid initial peak increase, remained elevated for a prolonged time period and then decayed non-linearly. In the presence of EGTA, the signal was smaller and showed an immediate linear decay. The peak Ca²⁺ component obtained in the presence of extracellular Ca²⁺ represents the change in Fura Red fluorescence due to Ca²⁺ entry. Although the mean peak increase in Fura Red fluorescence induced by 100 μM ATP in the presence of extracellular Ca²⁺ was significantly higher than that in the presence of EGTA, there was a remarkable variability of the data (Fig. 6C). Variability could theoretically depend on (1) subtle variations of the focal plane during recordings in TIRF mode; (2) different expression levels of juncate–YFP in the puncta, or (3) a combination of both. We excluded the first possibility because our TIRF set up was equipped with an on-line control of the focal plane, which avoids any focal drift, a crucial parameter during TIRF recordings. Since the fluorescence intensity of junctate–YFP puncta was quite heterogeneous (Fig. 6Ad), we analyzed the relationship between the peak Ca²⁺ signal increase and junctate–YFP fluorescence intensity (Fig. 6D,E). Although with some scatter, the data show a parallel increase in both parameters (Fig. 6D). Indeed, the relationship was linear with a good correlation (R=0.857) at low to intermediate juncate–YFP fluorescence levels (Fig. 6E), but the peak Ca²⁺ signal saturates in puncta exhibiting high levels of expression of juncate–YFP (Fig. 6D).

In the present work, we studied the structural and functional relationship of a supramolecular complex involved in regulating Ca²⁺ homeostasis of non-muscle cells. Our results show that junctate forms distinct puncta in regions on or very close to the PM, whose size and distribution are unaffected by mobilization of calcium from an InsP₃-sensitive store. Puncta are associated with the appearance of ER–PM peripheral junctions. Interestingly, such junctions are endowed with regularly spaced electron-dense structures that bridge the gap between ER and PM. We propose that these peripheral junctions contain junctate–InsP₃R–TRPC3 complexes because their number is dependent on the co-expression of protein components of the complexes, namely junctate and the Ca²⁺-entry channel TRPC3. Ca²⁺ imaging by TIRF microscopy demonstrates that puncta are domains in which agonist-activated Ca²⁺ entry occurs; the extent of Ca²⁺ entry correlates with the level of junctate expression.

An interesting observation arising from this study is that the N-terminus of junctate interacts with a domain on the InsP₃R3 involved in ligand binding (Maranto, 1994), making junctate a member of the group of proteins including IRBIT, RACK1, Homer, CaM, CaBP1 and TRPC3, which have been found to interact within residues 1–680 of the InsP₃R (Mikoshiba, 2007). The results of the pull-down assay are specific because no signals were detected in the Sepharose beads when we used an unrelated biotinylated peptide or when the pull down was carried out with a peptide derived from the N-terminus of aspartyl-β-hydroxylase/humbug, an alternative splice product of the gene encoding junctate, aspartyl-β-hydroxylase and junctin (Treves et al., 2009). The interaction of junctate with the InsP₃R is inhibited by InsP₃, the physiological ligand of the receptor. Such a result is reminiscent of that seen with IRBIT, the only other protein that binds to the InsP₃ binding domain on the InsP₃R to be displaced by InsP₃ (Ando et al., 2006). However, as opposed to IRBIT, junctate alters neither the K_D nor the B_max of InsP₃ binding, thus it does not act as a competitive inhibitor, but could be important in maintaining the structure of the InsP₃R macromolecular complex (Kiselyov et al., 2005) within the ER, a function similar to that attributed to 4.1N, another InsP₃R binding protein (Mikoshiba, 2007).

The ‘conformational coupling’ hypothesis, one of the models proposed to explain Ca²⁺ entry, predicts an interaction between Ca²⁺-entry channels and InsP₃Rs localized at ER–PM peripheral junctions (Irvine, 1990; Kiselyov et al., 2000; Berridge, 2004). Activation of such complexes may be triggered (1) by InsP₃ stimulation of a set of InsP₃R coupled to Ca²⁺-entry channels within peripheral junctions or (2) by opening perijunctional InsP₃Rs leading to localized depletion of peri-junctional ER stores and subsequent switching on of store-operated channels. The organization and assembly of ER–PM peripheral junctions are crucial aspects that still remain to be elucidated to prove the validity of this model. In the present report, we investigated the ultrastructural characteristics of HEK293 cells stably expressing junctate and junctate plus TRPC3. These HEK293 clones have a high number of peripheral ER–PM junctions that are characterized by the presence of electron-dense structures bridging the 10 nm gap between the ER and PM. Interestingly, the size of the junctional gap fits with measurements carried out using different experimental approaches, which show that in the junctional area the ER and PM are separated by 11–14 nm (Varnai et al., 2007; Orci et al., 2009). The ultrastructural properties of peripheral couplings in our HEK293 clones are strikingly similar to those of differentiating BC₃H1 cells, in which ordered arrays of Ca²⁺ release units made up of ryanodine receptors and L-type Ca²⁺ channels are present (Protasi et al., 1997). Similar peripheral couplings have been observed in developing myotubes (Protasi et al., 1997), in Xenopus oocytes (Gardner and Grey, 1983) and in a number of mammalian cells (Levine, 2004). In Xenopus oocytes, the contact sites have been postulated to be involved in Ca²⁺ influx (Gardner and Grey, 1983). Although we do not know the exact identity of the electron-dense structures present in the ER–PM peripheral couplings in HEK293 clones, we suggest that they are the structural expression of the upregulation of the components of the supramolecular complex made up by junctate–YFP and TRPC3. The fluorescence of junctate–YFP was exploited to visualize expression of the protein in HEK293 clones by TIRF microscopy. The appearance of peripheral couplings is associated with the formation of puncta that are on or very close to the PM. The link between the appearance of peripheral couplings and the expression of junctate–YFP strongly suggests that the fluorescent puncta are subcellular domains endowed with ER–PM peripheral junctions. At first sight, the fluorescent puncta might seem to be more numerous and larger than the EM peripheral couplings; however, it should be kept in mind that the latter technique only looks at very thin sections. As to the actual size of the fluorescent puncta, these might be smaller than calculated by TIRF because of the spread function of the microscope. In addition, the cells analyzed by TIRF microscopy are living and adherent to the cover glass, and ER–PM junctions might be more abundant on the basal membrane facing the glass support compared with those on the apical membrane (Orci et al., 2009). The cells processed for EM imaging were resuspended and probably lost the polarity of the ER–PM interaction.

In TIRF mode we placed the laser beam at the most external region of the rear focal plane of the objective. Under this condition, the calculated penetration depth of the evanescent wave is up to 60 nm. Thus, the fluorescent signal of the puncta is derived from the 60 nm space adjacent to the internal face of the plasma membrane and might also be due to junctate–YFP localized in peri-junctional ER membranes. Nevertheless, bearing in mind the limitation of the spatial resolution of the TIRF, we suggest that puncta containing junctate–YFP represent Ca²⁺-entry units that are probably different from those involved in SOCE induced by oligomerization of STIM1 with Orai1. The most important difference between the puncta induced by STIM1 and those containing junctate is that the appearance of the latter is not dependent on the level of Ca²⁺ load in the ER. We also did not observe any change in size or localization of junctate-containing puncta upon addition of an InsP₃R-mobilizing agonist. A possible explanation of these data in line with the EM structural data is that junctate stabilizes the associations between the ER and PM.

Ca²⁺ entry has recently been proposed to be secondary to the formation of ER–PM peripheral junctions mediated by the oligomerization of STIM1 and Orai1. However, this general scheme has been revisited on the basis of experiments demonstrating that activation of Ca²⁺ entry via Orai1 can occur in STIM1-deficient cells, and that the cytoplasmic domain of STIM1 is too small (Varnai et al., 2007) to bridge the 13-nm-wide junctional ER–PM gap (Orci et al., 2009). Our data suggest that Ca²⁺ entry can occur without de novo translocation of ER domains to subplasmalemma domains. Rather, agonist-activated Ca²⁺ entry can occur within junctate-containing puncta that do not change in size and location upon surface receptor activation of calcium release from the ER. The Ca²⁺ signal we recorded in TIRF mode contains a component derived from release from internal stores because it occurs in the presence of 0.5 mM EGTA. The additional Ca²⁺ signal component observed in the presence of extracellular Ca²⁺ is slower and must be derived from activation of Ca²⁺ entry. The latter component is probably mediated by a molecular complex modulated by junctate because it strongly correlates with the expression level of this protein in the puncta. In addition, we think that it is specific because it saturates at high levels of expression of junctate.

These results underlie the importance of junctate–InsP₃R–TRPC3 supramolecular complexes to form stable ER–plasma membrane peripheral couplings that might represent an anatomical site of agonist-activated Ca²⁺ entry.

Fragments of InsP₃R3 were prepared as described (Tang et al., 2001). To verify the size of the polypeptides and translation efficiency, aliquots of the ³⁵S-labelled translation mixture were resolved by 10% SDS-PAGE. Gels were stained with Coomassie Brilliant Blue (CBB), dried and exposed to autoradiography. For pull-down studies, streptavidin-coated magnetic beads were washed three times with 10 mM HEPES pH 7.2, 150 mM NaCl (HEPES-NaCl) and incubated for 30 minutes at room temperature with a 10% blocking solution (Roche Applied Science) in HEPES-NaCl. Beads were rinsed with HEPES-NaCl and incubated with 10 μg/ml biotinylated synthetic peptides corresponding to (1) the N-terminus of junctate (MAEDKETKHGGHKNGRK), (2) the N-terminus of aspartyl-β-hydroxylase (AQRKNAKSSGNSSSSGSTSAGSSSPGARF) and (3) an unrelated biotinylated peptide, for 30 minutes at room temperature. Beads were washed three times with HEPES-NaCl and divided into five samples. To each Eppendorf tube containing biotinylated peptide bound to streptavidin beads, 5 μl ³⁵S-labelled in vitro synthesized polypeptides of the InsP₃R (F1–F5) were added. After incubation for 1 hour at room temperature, beads were spun down and unbound material (void) was retained. Pellets were washed four times with 1 ml HEPES-NaCl, after which 40 μl of 5% SDS were added to the pellets; mixtures were incubated for 5 minutes at 95°C, spun and the supernatants (bound) were retained. Proteins in the samples corresponding to void, last wash and bound were separated by 10% SDS-PAGE; after CBB staining, gels were dried and processed for autoradiography.

InsP₃ concentration-dependent binding studies were carried out on HEK293 microsomes (75 μg) in 100 mM KCl, 10 mM NaCl, 10 mM Tris–HCl, pH 7.5, with [³H]Inositol-1,4,5-trisphosphate (Perkin Elmer; specific activity 0.001 mCi/ml) plus cold InsP₃ (Calbiochem) to give a final concentration of InsP₃ of 10, 25, 50, 100 and 150 nM, in the presence of 50 μM of an unrelated peptide or a synthetic peptide corresponding to the N-terminus of junctate. Values for non-specific binding were obtained by incubating microsomes with an excess (10 μM) unlabelled InsP₃. Samples were incubated for 30 minutes at 4°C, filtered through Whatman glass microfibre GFP filters by means of a Millipore Manifold filtering apparatus and the amount of radioactivity retained on the filters was determined by liquid scintillation counting.

HEK293 cells were transfected with the cDNA encoding junctate-EYFP (full-length junctate subcloned into the EYFP-C1 vector) using calcium phosphate. Stable clones (HEK junctate–YFP) were selected by adding G418 (final concentration 800 μg/ml) to the growth medium (DMEM + Earl's salts, containing 10% foetal bovine serum and penicillin-streptomycin). After 2 weeks in selective medium when they became visible, single clones were picked and expanded. For HEK clones overexpressing TRPC3 and junctate, the full-length junctate cDNA was subcloned into the HindIII site of the pcDNA 3.1/Hygro plasmid (Invitrogen life technologies). The construct was checked by sequencing and the plasmid was used to transfect a HEK293 cell line stably expressing TRPC3 (HEK T3 cells) (Zhu et al., 1998) using the ESCORT IV transfection reagent (Sigma, St Louis, MO) at a DNA: liposome ratio of 1:4, following the manufacturer's instructions. Stable clones (HEK T3–junctate) were selected by adding G418 and hygromycin (800 μg/ml and 200 μg/ml, respectively) to the growth medium (DMEM containing 10% foetal bovine serum, and penicillin-streptomycin). Three weeks after transfection, single clones were picked and expanded.

Total RNA was extracted from 2×10⁶ HEK293, HEK Junctate–YFP and clone HEK T3–junctate cells (Qiagen, Basel, Switzerland) and treated with Deoxyribonuclease I (DNaseI) (Invitrogen, Carlsbad, CA). After reverse transcription using 1000 ng RNA and the Moloney Murine Leukemia Virus Reverse Transcriptase (M-MLV RT) (Invitrogen), cDNA was amplified by quantitative real-time PCR in the ABI PrismTM7700 using the TaqMan® technology. Commercially available exon-intron junction-designed primers for human GAPDH and InsP₃R3 (Applied Biosystems, Foster City, CA) were used. Gene expression was normalized using self-GAPDH as reference and calculated as a fold increase compared with gene expression in HEK293 cells. In the case of junctate, real-time RT-PCR was carried out using SYBR Green technology (fast SYBR green master mix, Applied Biosystems) and the following primers: junctate forward, 5′-ATGGAGGACACAAGAATGGGA-3′ and reverse, 5′-CCAAACGACAGCTACAGATGT-3′. Gene expression was normalized using self-TATA box binding protein (TBP) as reference using the same technique and the following primers: forward, 5′-GCCATAAGGCATCATTGGAC-3′ and reverse, 5′-AACAACAGCCTGCCACCTTA-3′. For semiquantitative RT-PCR, total RNA was isolated from HEK cell clones using ULTRASPEC reagent (Biotecx) and transcribed into cDNA using a kit from Roche Diagnostics, oligo- (dT) as primers and AMV Reverse Transcriptase as previously described (Treves et al., 2004). RT PCR was carried out as previously described for junctate (Treves et al., 2004) and using the following primers and conditions for TRPC3: forward, 5′-ATG CTG CTT TTA CCA CTG TAG-3′ and reverse 5′-TCC TTC TGC ATT TGG GAA A-3′ (95°C for 5 minutes, followed by 35 cycles of 45 seconds 94°C, 45 seconds 51°C, 45 seconds 72°C; followed by extension at 72°C for 3 minutes); β-actin forward, 5′-TGA CGG GGT CAC CAC-3′ and reverse, 5′-CTA GAA GCA TTG CGG T-3′ (95°C for 5 minutes, followed by 25 cycles of 30 seconds 94°C, 45 seconds 49°C, 45 seconds 72° followed by extension at 72°C for 3 minutes). PCR products were separated and the DNA bands visualized by ethidium bromide staining.

SDS-PAGE, protein transfer onto nitrocellulose and immunostaining were performed as previously described (Treves et al., 2000; Treves et al., 2004). Blots were probed with a polyclonal goat anti-InsP₃R1, anti-InsP₃R2 and or anti-InsP₃R3 antibodies and anti-β-tubulin (Santa Cruz) or mouse anti-GFP (Roche Applied Science) followed by peroxidase-conjugated secondary antibodies. The immunopositive bands were visualized by chemiluminescence using the Super Signal West Dura kit from Thermo Scientific. For immunofluorescence, glass coverslip grown HEK293 junctate–YFP clone 21 cells were fixed with ice-cold acetone:methanol (1:1) for 20 minutes, rinsed twice with phosphate-buffered saline (PBS) and blocked with 10% blocking buffer (Roche Applied Science) for 60 minutes at room temperature. Coverslips were incubated with polyclonal goat anti-InsP₃R3 antibody (Santa Cruz); after 60 minutes, coverslips were rinsed three times for 5 minutes with PBS and incubated with rabbit anti-goat DyLighte Fluor-405 (400 nm excitation, 420 nm emission) (Rockland, Gilbertsville, PA) and rabbit anti-GFP Alexa Fluor 488. After 60 minutes, coverslips were extensively washed and mounted in glycerol mounting medium. Fluorescence was visualized by TIRF microscopy as described in the next section, by exciting at 488 nm (Sapphire Laser) to visualize GFP fluorescence, and at 405 nm using a laser from Coherent Laboratories. DyLighte Fluor-405 fluorescence was visualized using a BrightLine CH 427 filter (AHF Analysentechnir, Tübingen, Germany). Colocalization analysis of junctate and InsP₃R3 was performed using the colocalization application included in the Metamorph 5.7.4 software package.

For TIRF microscopy, HEK junctate–YFP stable clones or HEK293 transiently transfected with a plasmid encoding a GFP fusion protein targeted to the ER (SScalEGFP was a gift from Hans-Peter Hauri, Biozentrum, Basel University, Switzerland) were grown on a 0.17-mm-thick glass coverslip. Cells were imaged with a Nikon Eclipse TE2000-E fluorescent microscope equipped with an oil immersion CFI Plan Apochromat 100× TIRF objective (1.49 numerical aperture). Changes in fluorescence were detected by exciting at 488 nm and recording the emission at 510 nm via an electron multiplier C9100-13 Hamamatsu CCD camera. The focal plane corresponding to the glass coverslip cell adhesion site was selected using the SRIC filter; this focal plane was maintained throughout the experiment, using the Perfects Focus System. Cells were stimulated with Krebs-Ringer containing 2 mM Ca²⁺ for 5 seconds, followed by a 10 second stimulation with 100 μM ATP. Analysis of the size and shape of fluorescent puncta was performed with the Integrated morphometric analysis application included in the Metamorph 5.7.4 software package. Movement analysis was performed on the acquired stacks using Metamorph 7.5 software. To track movement, individual regions of interest were selected, subjected to a 2×2 low-pass filter, after which threshold values were set. Movement was tracked on selected fluorescent puncta using the ‘track object’ software included in the Metamorph 7.5 software package.

For electron microscopy, HEK293 cell clones were gently detached from tissue culture dishes, rinsed with PBS, fixed in 3% glutaraldehyde in PBS and pelleted. Cells were maintained in fixative for several days. Pellets were post-fixed in osmium tetraoxide, en-bloc ‘stained’ in uranyl acetate and embedded. Thin sections showed 1–200 profiles of cells per section, cut in random orientations. Sections were stained in solutions containing uranyl acetate and lead and examined in a Philips 410 electron microscope.

For measurement of cell populations, HEK cells were loaded with 5 μM Fura-2 AM in Krebs-Ringer buffer containing 2 mM external calcium. After 30 minutes at 37°C, cells were rinsed and changes in Fura-2 fluorescence were monitored by following the changes in fluorescence at 340 and 380 nm (excitation) and 515 nm (emission) with a Perkin Elmer LS6 spectrofluorimeter equipped with a magnetically controlled thermostat chamber. Cells (1×10⁶ cells/ml) were stimulated either with thapsigargin (400 nM) or with 100 μM ATP in Krebs-Ringer containing 0.5 mM EGTA; in some experiments 0.5 mM Ca²⁺ was added back to stimulate Ca²⁺ influx.

Single-cell ratiometric Fura-2 Ca²⁺ measurements were performed as previously described (Treves et al., 2004). For single-cell Ca²⁺ measurements by TIRF microscopy, junctate–YFP cells grown on glass coverslips were loaded with 5 μM Fura Red AM at 37°C for 30 minutes in Krebs-Ringer containing 2 mM Ca²⁺. Coverslips were mounted in a thermostat perfusion chamber and the focal plane corresponding to the glass coverslip and cell membrane contact site was identified using the Surface Refection Interference Contrast (SRIC) filter and fixed with the Perfect Focus System. YFP fluorescent puncta were visualized by illuminating the cells with a Sapphire laser and monitoring fluorescence at 510 nm; maintaining the same focal plane changes in the [Ca²⁺] measurements were monitored by illuminating the cells with a 405 nm laser and following the emission at 625–640 nm using a brightline HC 625/26 cut-off filter (AHF Analysentechnir). Images were acquired in TIRF mode every 100 mseconds using a Nikon Eclipse TE2000-E fluorescent microscope equipped with an oil immersion CFI Plan Apochromat 100× TIRF objective as described above.

Statistical analysis was performed using the Student's t-test for unpaired samples; means were considered statistically significant when P<0.05. Linear regression analysis and statistical analysis was performed using the statistics package included in the Origin 6.0 (Micorcal) software.

We would like to thank Olivier Pertz for help with the particle tracking, Serge Summermatter for help with the real-time RT-PCR, Anne-Sylvie Monnet for expert technical assistance and Nosta Glaser for electron microscopy support, as well as Alex Eberle for the bitotinylated unrelated peptide. This work was supported by grants from Ministero Universita e Ricerca scientifica e Tecnologica 40%, F.I.R.B. RBAUO01ERMX, by the Department of Anaesthesia, Basel University Hospital and by National Institutes of Health grants NIH RO1 HL48093 to C.F.A. and ROI DK081654 to M.X.Z. Deposited in PMC for release after 12 months.