The role of the store‐operated calcium entry channel Orai1 in cultured rat hippocampal synapse formation and plasticity

Abstract

Key points

• The role of non‐synaptic calcium entry in the formation and functions of dendritic spines was studied in dissociated cultured rat hippocampal neurons.

• Orai1, a store‐operated calcium channel, is found in dendritic spines.

• Orai1 co‐localizes in dendritic spines with STIM2 under conditions of lower [Ca²⁺]o.

• Orai1 channels are associated with the formation of new dendritic spines in response to elevated [Ca²⁺]o.

• Lack of Orai1, either by transfection with a dominant negative construct or with small interfering RNA to Orai1, results in retarded dendritic spines, an increase in density of filopodia, lower synaptic connectivity and the ability to undergo plastic changes.

• These results highlight a novel role for Orai1 in synapse formation, maturation and plasticity.

Abstract

The possible role of store operated calcium entry (SOCE) through the Orai1 channel in the formation and functions of dendritic spines was studied in cultured hippocampal neurons. In calcium store‐depleted neurons, a transient elevation of extracellular calcium concentration ([Ca²⁺]_o) caused a rise in [Ca²⁺]_i that was mediated by activation of the SOCE. The store depletion resulted in an increase in stromal interacting molecule 2 (an endoplasmic calcium sensor) association with Orai1 in dendritic spines. The response to the rise in [Ca²⁺]_o was larger in spines endowed with a cluster of Orai1 molecules than in spines devoid of Orai1. Transfection of neurons with a dominant negative Orai1 resulted in retarded maturation of dendritic spines, a reduction in synaptic connectivity with afferent neurons and a reduction in the ability to undergo morphological changes following induction of chemical long‐term potentiation. Similarly, small interfering RNA (siRNA)‐treated neurons had fewer mature dendritic spines, and lower rates of mEPSCs compared to scrambled control siRNA‐treated neurons. Thus, influx of calcium through Orai1 channels facilitates the maturation of dendritic spines and the formation of functional synapses in central neurons.

Key points

• The role of non‐synaptic calcium entry in the formation and functions of dendritic spines was studied in dissociated cultured rat hippocampal neurons.

• Orai1, a store‐operated calcium channel, is found in dendritic spines.

• Orai1 co‐localizes in dendritic spines with STIM2 under conditions of lower [Ca²⁺]o.

• Orai1 channels are associated with the formation of new dendritic spines in response to elevated [Ca²⁺]o.

• Lack of Orai1, either by transfection with a dominant negative construct or with small interfering RNA to Orai1, results in retarded dendritic spines, an increase in density of filopodia, lower synaptic connectivity and the ability to undergo plastic changes.

• These results highlight a novel role for Orai1 in synapse formation, maturation and plasticity.

Abbreviations

CPA

cyclopiazonic acid

GFP

green fluorescent protein

LTP

long‐term potentiation

mEPSC

miniature EPSC

siRNA

small interfering RNA

SOCE

store operated calcium entry

STIM

stromal interacting molecule

SY

synaptophysin

Introduction

Calcium stores play a pivotal role in the regulation of the intracellular calcium concentration ([Ca²⁺]_i) in both neurons and non‐neuronal cells (Verkhratsky, 2005; Zalk et al. 2007). The release of calcium from stores is controlled by several mechanisms, which activate ryanodine and inositol trisphosphate receptors. Depletion of calcium from stores is sensed by the stromal interaction molecule (STIM). It clusters near the depleted store and relocates to the membrane, where it interacts with the Orai1 plasma membrane calcium channel to allow calcium influx into the cell, thus refilling the stores (Bogeski et al. 2012). The interaction of stores/STIM/Orai has been studied extensively in non‐neuronal cells, and its malfunction has been implicated in immunological diseases, including severe combined immunodeficiency syndrome (Feske et al. 2006). Compared to the vast literature available with respect to the functions of STIM/Orai1 in non‐excitable cells, relatively less is known about their role in central neurons. STIM and Orai are localized in the brain (Skibinska‐Kijek et al. 2009) and STIM1/Orai1 can be converted from a dispersed to a punctate appearance upon depletion of calcium stores with thapsigargin (Klejman et al. 2009). Both are important in the regulation of both growth cone motility (Mitchell et al. 2012) and voltage‐gated calcium channels (Park et al. 2010), as well as with respect to the detrimental effects of chronic epilepsy (Steinbeck et al. 2011) and oxidative stress (Henke et al. 2013). Earlier work ascribed a role for store operated calcium entry (SOCE) in synaptic plasticity, in that SOCE antagonists reduced long‐term potentiation (LTP) in hippocampal neurons (Baba et al. 2003). Furthermore, interacting proteins such as SARAF (Palty et al. 2012) and septins (Sharma et al. 2013; Tada et al. 2007) regulate SOCE, emphasising the important role of STIM/Orai1 in controlling [Ca²⁺]_i. The recent association of septins (Sharma et al. 2013; Tada et al. 2007) with SOCE is intriguing because septins have been found in dendritic spines of central neurons (Xie et al. 2007) and may provide a link between dendritic spines and calcium stores. Although STIM1 has been imaged (in punctate form) in dendrites of hippocampal neurons (Keil et al. 2010), the dynamic distribution and function of Orai1 in these neurons has remained enigmatic.

An unsettled issue is related to the association of Orai1 with STIM. There are two species of STIM in central neurons: STIM1 and STIM2. Although most studies associate STIM1 with Orai1 (Skibinska‐Kije et al. 2009), there are more recent studies indicating that STIM2 is the dominant species in hippocampal neurons (Sun et al. 2014; Zhang et al. 2015). Is either STIM1 or STIM2 associated with Orai1? Doe Orai1 or Orai2 represent the SOCE channel in hippocampal neurons?

In the present study, we combined time‐lapse imaging with molecular and electrophysiological tools to explore the role of Orai1 in cultured hippocampal neurons. We report that: (i) dendritic spines possess significant calcium stores; (ii) Orai1 is an activator of SOCE; (iii) upon depletion of calcium, STIM2 becomes co‐localized with Orai1 in the spines; and (iv) the presence of clusters of Orai1 molecules (= puncta) is correlated with the emergence of nascent spines on dendrites following a transient rise of [Ca²⁺]_o. Thus, we propose that Orai1 has an important role in the formation, maturation and plasticity of dendritic spines in developing central neurons.

Methods

Cultures

Animal handling was conducted in accordance with the guidelines published by the Institutional Animal Care and Use Committee of The Weizmann Institute and with the Israeli National guidelines on animal care. Cultures were prepared as detailed elsewhere (Korkotian and Segal, 2006; Vlachos et al. 2009). Briefly, rat pups were decapitated on the day of birth (P0–1), their brains removed, and the hippocampi were dissected free and placed in a chilled (4°C), oxygenated Leibovitz L15 medium (Gibco, Gaithersburg, MD, USA) enriched with 0.6% glucose and gentamicin (20 μg ml⁻¹; Sigma, St Louis, MO, USA). Approximately 10⁵ cells in 1 ml of medium were plated in each well, onto a hippocampal glial feeder layer that was grown on the glass for 2 weeks prior to plating of the neurons. Cells were left to grow in the incubator at 37°C with 5% CO₂. Neurons were transfected with Orai1‐GFP, dominant negative Orai1 or constitutive active Orai1 [the CA variant was G98D and the dominant negative variant was G98A (Zhang et al., 2011), a gift from Dr E. Reuveny, The Weizmann Institute, Rehovot, Israel] and DsRed or EBFP2 (to image cell morphology), using lipofectamine 2000, at 6–7 days in vitro (DIV) and were used for imaging at 10–21 DIV.

ORAI‐1 small interfering RNA (siRNA) transfection was performed using ON‐TARGET plus SMART pool siRNA (100 nm final concentration) (Dharmacon, Lafayette, CO, USA), comprising a pool of four siRNA sequences: GCCAUAAGACGGACCGACA, UCAAGAGGCAGGCGGGACA, CAACAGCAAUCCGGAGCUU and CCUGUGGCCUGGUGUUUAU. A non‐targeting pool of four RNA sequences was used as a control: UGGUUUACAUGUCGACUAA, UGGUUUACAUGUUGUGUGA, UGGUUUACAUGUUUUCUGA and UGGUUUACAUGUUUUCCUA. DharmaFECT1 transfection reagent (catalogue number T‐2001‐01) was used (1 ml well^–1) in accordance with the manufacturer's instructions. Transfection efficiency was tested using siGlo‐cre transfection indicator (Dharmacon) and found to penetrate 96% of the neurons. The cultures were exposed to the siRNA for 2–3 days before being used for imaging or electrophysiology. Because DarmaFect1 has an effect of its own with respect to increasing spontaneous activity, comparisons were always made between T and NT neurons in the same experiment.

Live cell imaging

Cultures were placed in the imaging chamber, controlled by an automated X‐Y stage (Luigs and Neumann, Ratingen, Germany). Neurons were imaged on the stage of an upright PASCAL confocal microscope (Carl Zeiss, Oberkochen, Germany) using an 63× water immersion objective (0.9 NA) (Olympus, Tokyo, Japan) and 2–4× zoom scans. Standard recording medium contained (in mm); 129 NaCl, 4 KCl, 1 MgCl₂, 2 CaCl₂, 10 glucose, 10 Hepes, with the pH adjusted to 7.4 with NaOH, and osmolality to 320 mosmol kg^–1 with sucrose. In some recording sessions, CaCl₂ was replaced by MgCl₂. TTX (0.5 μm) was added to the recording medium in some experiments to avoid indirect effects as a result of changes in network activity. Fluo‐2AM (2 μm; Invitrogen, Carlsbad, CA, USA) was incubated for 1 h at room temperature to image variations in [Ca²⁺]_i resulting from network activity. Alternatively, K⁺Fluo‐4 salt solution was injected into neurons with sharp micropipettes and allowed to diffuse for 0.5 h before imaging. Calcium (5 mm) containing medium was applied by pressure from the tip of a patch pipette (diameter 1–2 μm), near the imaged dendrite. In the earlier experiments (Fig. 1), 50 mm calcium was applied by pressure from a pipette tip. This concentration did not produce a major effect under control conditions (Fig. 1 D) but, to avoid possible artefacts, we used 5 mm CaCl₂ in the pressure pipette in subsequent experiments.

Figure 1. Store‐dependent rise in [Ca²⁺]_i following application of CaCl₂ through a pressure pipette.

A, configuration of the recorded neuron and the placement of a pressure pipette. A neuron was injected with Fluo‐4 through a sharp micropipette, allowed to rest for at least 30 min, and imaged on the confocal microscope. The fluorescence image is merged with phase image to visualize the pressure pipette. B, a close‐up view of a dendrite and adjacent spines, and an arrowhead showing where the line scan is taken of the dendrite and the spine. C, sample illustrations of line scans of the same dendrite taken under 0 Ca²⁺ conditions (left, illustrated in E) and under store‐depletion conditions (C, right, and F). The arrowhead denotes the time when pressure was applied. D, averaged line scans of 28 spine/dendrite pairs recorded under standard conditions (2 mm CaCl₂, 1 mm MgCl₂, in the presence of TTX). A medium containing 50 mm CaCl₂ was applied for the duration of the horizontal bar. E, same sample of dendrites/spines pairs in calcium‐free medium. A larger response to the pressure application of calcium is seen, especially in the spine compartment. F, same dendrites/spines after exposure to 15 μm CPA for 5 min, which was extensively washed out before the imaging session. G, summary diagrams of the change in fluorescence (expressed as Δf/f) in response to 50 mm CaCl₂, under the different conditions averaged in 28 pairs of dendrite/spines. The right bars are taken after CPA, in the presence of 30 μm 2‐APB, which blocks entry of calcium through SOCE channels (n = 19 pairs) and reduces the responses significantly. H, net responses (Δf/f) of nine dendrites in three different experiments to puff application of 5 mm CaCl₂ in standard 0 mm CaCl₂ (left column) and in the same dendrites, after wash‐in of TTX (1 μm), nifedipine (10 μm) and APV (50 μm). Addition of these drugs did not change the intracellular rise of [Ca²⁺] in responses to the puff of CaCl₂. [Colour figure can be viewed at wileyonlinelibrary.com]

Chemical LTP was induced in the cultured neurons by exposing them, for 3 min, to a NMDA receptor enhancing medium containing 3 mm CaCl₂, 0 mm MgCl₂ and 100 μm glycine. This resulted in a massive enhancement of synaptic activity (Korkotian et al. 2014). The medium was then replaced by the standard recording medium and the same cell/dendrite was imaged 60 and 90 min later.

Electrophysiology

Recording was made under standard conditions, using patch pipettes, as described previously (Vlachos et al. 2009). Signals were amplified using a Multiclamp 700 B amplifier, accumulated and analysed with Pclamp, version 9 (Molecular Devices, Sunnyvale, CA, USA) and MiniAnalysis software (Synaptosoft Inc., Fort Lee, NJ, USA). Miniature EPSC (mEPSCs) were recorded in the presence of TTX (0.5 μm) and bicuculline (10 μm). In some experiments, a puff of high osmolality medium (addition of 300 mm sucrose to the standard medium) was applied through the tip of a pipette near the recorded neuron to produce an increase in the rate of mEPSCs.

Immunostaining

Cover glasses bearing transfected primary hippocampal cells were washed briefly with standard extracellular solution. Cultures were then fixed with 4% paraformaldehyde in 0.1 m PBS (pH 7.4) for 20 min, and subsequently washed with PBS thoroughly. Cultures were incubated for 1 h with 10% normal horse serum in 0.1% Triton X‐100 containing PBS and subsequently incubated for 24 h at 4°C in anti‐Orai1 antibody (D‐15, goat polyclonal, dilution 1:200 in PBS; Santa Cruz Biotechnology, Santa Cruz, CA, USA; or mouse monoclonal (266.1) Ab, 1 mg ml⁻¹; Abcam, Cambridge, MA, USA). The two antibodies identified the same Orai1 puncta, with the antibody from Santa Cruz Biotechnology being more specific (data not shown). Anti Orai2 (rabbit, dilution 1:200; Santa Cruz Biotechnology), anti STIM1 (rabbit, dilution 1:200; Santa Cruz Biotechnology), anti STIM2 (goat polyclonal, dilution 1:200; Santa Cruz Biotechnology) or anti‐synaptophysin (MAB 5258, dilution 1:1000 in PBS to stain presynaptic terminal; Millipore, Billerica, MA, USA) were used in combinations under different testing conditions. Cultures were incubated for 1 h with Alexa 568‐labelled or Alexa 633‐labelled anti‐goat or anti mouse secondary antibody (dilution 1:200 in PBS; Molecular Probes, Eugene, OR, USA). Cover slips were rinsed, transferred onto glass slides and mounted for visualization on a LSM 510 or LSM 880 (Carl Zeiss) (which allows simultaneous visualization of four fluophores) with anti‐fading mounting medium. In all cases, secondary and tertiary dendritic segments were visualized. Confocal image stacks were used for 3D reconstructions. Images were prepared using Photoshop CS4 (Adobe Systems Inc., San Jose, CA, USA).

Statistical analysis

Fluorescence intensity was calculated using ImageJ (NIH, Bethesda, MD, USA) and Matlab R2010b‐based linescan acquisition software (MathWorks Inc., Natick, MA, USA). Measurements were made in a double‐blind procedure to ensure unbiased observations. Dendritic protrusions were categorized into mushroom spines, which consist of a head larger than 0.5 μm in diameter; thin and long spines, which are longer than 1.5 μm, with a distinct but small head that is thicker than the neck; stubby/short spines, where a neck could not be clearly visualized; and filopodia, which are devoid of a discernable head. In some experiments, the two categories of spines (mushroom and long) were merged. Some of the morphological analysis was conducted by three independent observers. Dendritic spines that were used for calcium imaging were identified in the immunostained or DsRed‐transfected neurons, and analysed independently of the measurements of calcium transients in these same spines. Statistical comparisons were made with t tests or ANOVA, as appropriate, using Matlab, KaleidaGraph (Synergy Software, Reading, PA, USA) and Origin software (OriginLab Co., Northampton, MA, USA).

Results

SOCE in dendritic spines

In the initial series of experiments, we confirmed the role of SOCE in loading depleted calcium stores in cultured rat hippocampal neurons. This was evaluated by the responses of hippocampal dendrites and spines to pressure application of 50 mm CaCl₂ under different testing conditions (Fig. 1). Individual cultured neurons were microinjected with Fluo‐4 to measure variations in [Ca²⁺]_i. In standard recording medium containing 2 mm CaCl₂, pressure application of CaCl₂ produced only a minor rise of [Ca²⁺]_i (Fig. 1 A and D). When the same cultured neurons were deprived of [Ca²⁺]_o, pressure application of CaCl₂ produced a slightly larger transient rise of [Ca²⁺]_i (Fig. 1 C, left, and E). Strikingly, a short (5 min) preincubation with 15 μm cyclopiazonic acid (CPA), a calcium uptake inhibitor that depletes calcium from stores, produced a fast and marked increase of [Ca²⁺]_i in response to the same amount of CaCl₂ (Fig. 1 C, right, and F). Interestingly, under all the three conditions tested, the response to a rise in [Ca²⁺]_o was slightly higher in the spine than in the parent dendrite compartment, as measured simultaneously using fast line scan across spine/dendrite segment (n = 28 dendrite/spine pairs, difference not significant statistically) (Fig. 1 G). In addition, the SOCE antagonist 2‐APB (at 30 μm) suppressed significantly the response of both spines and dendrites to the calcium load (n = 19 pairs, P < 0.01) (Fig. 1 G). These results indicate that calcium crosses the membrane, under conditions where the cell is deprived of stored calcium, using a SOCE channel, which probably resides in the membrane of both spines and dendritic shafts.

Finally, we examined the possible contribution of voltage and NMDA‐gated calcium channels to the rise of [Ca²⁺]_i following exposure to high CaCl₂ in the recording medium. In nine different dendrites (three experiments), we exposed neurons to calcium free medium for a prolonged time (30 min) and recorded their responses to puffs of CaCl₂, as above. We then added TTX (1 μm), nifedipine (10 μm) and APV (50 μm) and recorded the responses to a puff of CaCl₂ (Fig. 1 H). There was no difference in the [Ca²⁺]_i rise in response to the puff of calcium, indicating that either voltage or NMDA‐gated currents do not contribute to the observed rise in [Ca²⁺]_i.

The SOCE channels involved in the calcium surge

Cultures were exposed to the standard recording medium containing 2 mm CaCl₂, to calcium‐free medium for 30 min and to calcium‐free medium followed by replacement of calcium back into the medium. Each of the culture glasses contained green fluorescent protein (GFP) transfected neurons to enable visualization their dendrites and spines. Cultures were then fixed and stained with a combination of three out of four antibodies to stain for Orai1, Orai2, STIM1 and STIM2. The objective of this experiment was to determine whether any of the four antibodies co‐localize in dendritic spines under basal conditions, and in the nominal absence of [Ca²⁺]_o. Each of the analysed groups of neurons (three groups, including control, 0 Ca²⁺ and back to control) consisted of 12 GFP transfected neurons and, in each neuron, 28 randomly selected spines, without prior knowledge of their immunostaining, were analysed for the presence of the respected antibodies. Although Orai1 had a punctate distribution, and was present in a large proportion of dendritic spines (see also below) (Zhang et al. 2014), Orai2 was detected in more dispersed form, and mainly in dendritic shafts but to a lesser extent in spines (Fig. 2 B). Following removal of CaCl₂ from the incubation medium, there was no clear re‐distribution of Orai1 or Orai2 in the dendritic shaft/spines compartments. By contrast, although STIM1/2 were dispersed in dendrites under the control conditions, following removal of extracellular calcium, STIM2 emerged in punctate form in dendritic spines, significantly co‐localized with Orai1 puncta (Fig. 2 A, D and E). Surprisingly, STIM1 was much less mobile in the absence of [Ca²⁺]_o and appeared to move only into a subpopulation of spines where both STIM2/Orai1 were found (Fig. 2 C, top: Orai1/STIM2; bottom image of the same dendritic segment: STIM1; and E, right‐hand bars, and F, left‐hand bars, where Orai1 hardly co‐localizes with STIM1, in the absence of STIM2). Thus, our results indicate that STIM2/Orai1 is the dominant configuration associated with the activation of SOCE in dendritic spines.

Figure 2. Colocalization of STIM2 and Orai1 in dendritic spines upon depletion of [Ca2+]o.

A and B, dendritic segments immunostained for Orai1, Orai2, STIM1 and STIM2, on a green background of GFP‐transfected neurons. Each of the imaged dendrites is stained with two antibodies, as indicated in colour. Different dendritic segments from different cells are represented by two testing conditions. C, the same dendritic segment in the top and bottom frames, stained for Orai1, STIM1 and STIM2. The staining is shown in different images of the same dendrite for simplicity. Spines that contain STIM1 (bottom) also contain STIM2 and Orai1 (top). D, summary of the density of puncta detected in mushroom dendritic spines in control (red bars) in the absence of calcium (blue) and following replacement of calcium (green). The bars represent the percentage of labelled spines with a single protein (including double‐ and triple‐labelled spines). The only highly significant change in staining density was that of STIM2, which then returned to baseline following replacement of calcium. E, co‐localization of Orai1 and STIM2 (left) and Orai2 and STIM2 (middle) and Orai1/STIM1/STIM2 (right). There was a highly significant increase in Orai1/STIM2 under the zero calcium condition that was not mirrored in the Orai2 group. F, net number of co‐localized puncta, showing four different combinations of Orai and STIM, under conditions where the other peptide is not detected (e.g. Orai1/STIM1, in spines where STIM2 is not found). As shown before, the only highly significant co‐localized puncta in spines was that of Orai1 and STIM2 (n = 28 spines in each of 12 neurons, counted as percent of total number of mushroom spines). Significance levels in (D) to (F): ^* P < 0.05, ^** P < 0.01, ^*** P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com]

Orai1 in dendritic spines

Orai1 has been shown to constitute a transmembrane calcium channel to which the calcium sensor STIM binds to allow influx of calcium ions. Orai1 is distributed in discrete puncta unevenly along the plasma membrane. To correlate the presence of Orai1 puncta with the reactivity to the calcium puffs, we used neurons that were transfected with DsRed to visualize the morphology of the cell. The cultures were pre‐loaded with Fluo‐2AM. We applied 5 mm CaCl₂ prepared in normal recording medium from a patch pipette placed near dendritic spines/shafts of the imaged neuron (Fig. 3 A, lower). The responses to the calcium puffs were imaged in a line scan mode, comprising spines and their parent dendrites (Fig. 3 C and D). Subsequently, the imaged cultures were fixed and immunostained for the presence of Orai1 in spines/dendrites, and for the presynaptic marker synaptophysin (SY) (Fig. 3 B–D). Following positive identification of the imaged neurons, the responses to calcium application were compared between the different dendrites/spines. There was a significant difference in the magnitude of the [Ca²⁺]_i surge between Orai1(+) and Orai1(–) spines (mean = 0.99 ± 0.11, n = 25, and 0.26 ± 0.06 respectively, n = 30, p < 0.0001) (Fig. 3 E). These experiments were conducted both with 1‐ and 3‐week‐old cultures. Interestingly, although most of the spines (63.1 ± 8.9%) in young cultures are endowed with Orai1 puncta but missing SY, (Fig. 3 F), only ∼40% of them contained Orai1 at 3 weeks of age, at which time most of them were already touched by a SY containing terminal (Fig. 3 F). The difference between Orai1(+) and Orai1(–) spines spanned across the two age groups (data for comparison between ages not shown).

Figure 3. Orai1(+) and Orai1(–) spines differ in response to local pressure application of CaCl₂ .

A, low power view of a DsRed‐transfected neuron. Below is a zoomed image of a dendrite and the position of the pressure pipette next to a spine (on the left). B, the same neuron, after fixation and immunostaining for Orai1 (green) and synaptophysin (SY, blue). C and D, illustrations of two experiments with dendrites and spines responding to pressure application of 50 mm CaCl₂. Top image is line scans across a dendrite and two spines, on opposite sides of the dendrite. The same dendrite was immunostained and the presence of Orai1 and SY was detected in the left spine (marked with a red dot). Below is a summed line scan of the three regions of interest (note the dot colours matching those of the graph). The left spine has the largest response to CaCl₂ compared to the response of the parent dendrite and the right spine, which do not stain for Orai1. D, a similar illustration of a large spine, found in post hoc immunostaining to contain Orai1 and touched by SY, responding to pressure application of 50 mm CaCl₂, with a much larger response than that of its parent dendrite. E, summary diagram of the distribution of the magnitudes of the responses to CaCl₂ in Orai1(+) and Orai1(–) spines in Δf/f units. A significant difference between the groups is obvious (see text for details). F, distribution of Orai1 (left) and SY‐positive (right) spines at two age groups, 1 week (left bar) and 3 weeks (right bar) in culture. The large majority of spines in the young neurons contained Orai1 but did not contain SY, whereas, at the more mature age, most of them were endowed with SY puncta, although a smaller proportion of the spines contained Orai1. Significance levels: ^* P < 0.05, ^*** P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com]

To obtain a better control of the distribution and quantity of Orai1 in the dendrites and spines, we transfected 1‐week‐old cultured hippocampal neurons with one of the following plasmids: Orai1‐cherry, Dominant negative (DN) Orai1, DN together with Orai1‐cherry (Rescue, DN/R) and Constitutive active (CA) Orai1. Initially, we compared, in the same set of Orai1‐transfected neurons, spines that were endowed with Orai1 puncta [Orai1(+)], with those that were Orai1(–). There was a striking difference between Orai1(+) and Orai1(–) spines/dendrites in the same neurons, in response to a puff of calcium (Δf/f = 1.78 ± 0.17, n = 41 and 0.36 ± 0.12, respectively, n = 58, P < 0.0001) (Fig. 4 A–C). Interestingly, the [Ca²⁺]_i surge was larger in spines that were endowed with Orai1 than in spines that did not contain it, although, where there was an Orai1 punctum also in the parent dendrite, their responses matched those seen in the spine (Fig. 4 A and B). These results indicate that the transfected species of Orai1 may reflect the native one, in that it supports the influx of calcium through the SOCE mechanism.

Figure 4. Transfection with different Orai1 species alters responses to pressure application of CaCl₂ .

A and B, line scans in Fluo‐2 loaded neurons, transfected with Orai1; red puncta detected in spine and dendrite (A) or only in dendrite (B). C, comparisons of the magnitude of the response (in Δf/f) between Orai1(+) and Orai1(–) spines in the same populations of neurons. D and G, responses of spine/dendrite pairs to two successive applications of CaCl₂ in spine/dendrite pairs of neurons loaded with Fluo‐2AM and transfected with control (D), dominant negative (DN; E), DN+Orai1 for rescue (F) and constitutively active (CA; G) Orai1. (The responses are smaller than those measured in (C) because they did not distinguish between Orai1(+) and Orai1(–) spines in this experiment). H, magnitudes of responses in Δf/f in the four groups of transfected neurons, for both dendrites and spines. I, the responses in the control dendrites and spines are significantly suppressed by 2‐APB, a SOCE antagonist, as in Fig. 1(G). Significance levels in (H) and (I): ^* P < 0.05, ^** P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com]

To test this assumption more directly, comparisons were made between dendrites/spines of the four groups of neurons transfected with the different species of Orai1 (Fig. 4 D–I). Cultures were preincubated with Fluo‐2AM prior to the experiment, to visualize transient changes in [Ca²⁺]_i. In these experiments, we scanned small frames rather than lines, aiming to increase the spatial resolution at the same time as reducing temporal resolution. Two consecutive puffs of 50 mm CaCl₂ were applied near the spine/shaft segment in the field of view. The CA group of spines (n = 28 spine/dendrite pairs) produced significantly larger [Ca²⁺]_i surges than those of the control group, whereas the DN group (n = 36) generated significantly smaller [Ca²⁺]_i response than the controls. The rescue group (n = 18 pairs) generated larger responses than the DN ones, although this still did not recover to the level of the control (n = 44) group (Fig. 4 H). On the other hand, the time to peak was fastest in the CA group, whereas the DN group was slowest to reach the peak response (9.4 ± 0.57 s in control, 11.7 ± 0.93 s in DN, 8.8 ± 0.74 s in CA and 11.2 ± 0.75 s in the four groups of spines). Finally, the responses to the rise in [Ca²⁺]_o were suppressed by the presence of 30 μm 2‐APB, a SOCE antagonist (Fig. 4 I). These experiments indicate that Orai1 may have a pivotal role in influx of calcium into dendritic spines and their parent dendrites.

Orai1, spine formation and plasticity

The functional relevance of these observations was studied using time‐lapse imaging of dendritic segments following exposure of neurons to local pressure application of 5 mm CaCl₂ from a nearby patch pipette in the field of view of the imaged neurons at 10 or 20 DIV. Cultures were exposed to TTX and there was no calcium in the recording medium. 3D reconstructed images were taken at different times before and after the transient exposure to 5 mm CaCl₂ (Fig. 5 A). A cohort of 41 fields was exposed to CaCl₂ (Fig. 5 A and B) and the results indicate a highly significant (P < 0.01) difference between Orai(+) and Orai(–) spines. Furthermore, when the spines were further away from the pressure pipette, there was no difference between the two groups. Yet another group of dendrites (in 66 fields) was examined at the age of 20 DIV, in response to 5 mm CaCl₂ (Fig. 5 B). Strikingly, exposure to calcium did not produce novel spines in the 20‐day‐old neurons, regardless of the presence of Orai puncta. This difference is exemplified in a cumulative plot of new spines in the imaged neurons (Fig. 5 C and D).

Figure 5. Time lapse imaging of neurons transfected with GFP for morphology and Orai1 (red).

A, before (top) and following exposure to a brief rise in [Ca²⁺]_o. Several new dendritic spines are formed in the field of view over a period of 20 min. They were classified as Orai1(+) (68 of 97; 70% new spines) and Orai1(–) (29 of 97; 30%) spines. B, summary of new spine densities in 41 fields following exposure to 5 mm [Ca]_o, divided into Orai(+) and Orai(–) spines, at 10 and 20 DIV. Dendrites that were far apart from the pressure pipette (middle columns, apart) generated fewer new spines over the inspection interval. Significance levels ^** P < 0.01. C and D, cumulative histograms of the formation of new spines following exposure to 5mM [Ca²⁺]_o in 10 DIV (D) and 20 DIV (D) neurons, demonstrating the faster formation of new spines, endowed with Orai puncta. In this case, 97 new Orai1(+) spines were detected as well as 46 Orai(–) spines, although the latter group emerged slower than the first one in time after the exposure to calcium. In 20 DIV neurons (D) the Orai1(+) spines appear at about the same rate as the Orai1(–) ones over time. [Colour figure can be viewed at wileyonlinelibrary.com]

To examine the long‐term role of Orai1 in formation of dendritic spines and synaptic connectivity, we quantified the spine density/shape in the four groups of transfected neurons (GFP control, DN, DN+Rescue and CA) (Fig. 6). Transfection was performed at 6 or 7 days and fixation at 12–14 days. There was a significant difference in spine shape among the four groups. Although the majority of protrusions in the Orai1/GFP group (n = 54 fields), as well as the CA/GFP group (n = 48 fields), were of the mature, mushroom or short/stubby types (Fig. 6 A and B), the majority of the protrusions in the DN group (n = 87 fields) were of the immature, elongated, filopodia‐like structure. As before (Fig. 4), the rescue group (n = 31 fields) was between the control and the DN groups. These differences were highly significant (ANOVA for the mushroom spines, F = 11.86, P < 0.004, all Tukey comparisons were highly significant, except for the comparison between CA and Ct groups; ANOVA for the filopodia, F = 17.19, P < 0.001, with highly significant differences between DN and CA, DN and control, and DN and rescue).

Figure 6. Transfection with different Orai1 species affects dendritic spine morphology.

A, sample illustrations of different dendrites of neurons transfected with GFP + the different constructs. B, distribution of different types of dendritic protrusions, divided into mushroom type (M), long mushroom (L), small/stubby spines (S) and filopodia (F). The measurements were made in 54 fields (for the controls) 87 fields for the DN cells, 31 fields for the DN+R cells and 74 fields for the CA species. C, spine length in the four populations of spines, demonstrating that DN are the longest (reflecting the fact that most of them are filopodia). D, transfected neurons were immunostained for synaptophysin (red dots) and the numbers of red puncta touching dendrites and the spines of the transfected neurons were counted for each 100 μm length of dendrite. E, summary of the measurements in the different neuronal groups, showing that DN cells have significantly fewer SY puncta attached to them than the controls or the CA neurons. F and I, miniature excitatory postsynaptic currents (mEPSCs) recorded from neurons of the four transfected species. F, sample illustration of current traces, recorded from control (left) and DN (right), and an expanded trace of a single event, in the middle. Cells were clamped at −60mV. G, averaged current magnitudes recorded from the neurons in the four groups (Control, n = 14, DN, n = 11, CA, n = 20 and Rescue, n = 10). There were no differences among the groups in mEPSC amplitudes. H, frequency of mEPSCs in the different groups of neurons. The DN had significantly fewer events in a standard 2 min recording time than the control and CA groups. I, rise time of the mEPSCs is not different among the groups. [Colour figure can be viewed at wileyonlinelibrary.com]

An indication of maturity of the synaptic connections on the dendritic spines is the presence of presynaptic terminals stained with SY puncta attached to them. As shown before, control (n = 17 fields) and CA (n = 14 fields) groups had a significantly larger proportion of spines endowed with a presynaptic terminal compared to the DN (n = 29 fields) and the Rescue groups (n = 11 fields) (Fig. 6 D and E).

To determine whether the differences in maturity of the dendritic spines have any functional relevance, we recorded spontaneous mEPSCs from neurons in the four groups: control (n = 14 cells), DN (n = 11), CA (n = 20) and rescue (n = 10) neurons (Fig. 6 F–I). The neurons were imaged before recording to confirm that they are fluorescent, as were those that were analysed morphologically. Although there were no differences among the four groups tested in amplitudes of the mEPSCs, there was a significant reduction in frequency of mEPSC in the DN group compared to control and CA groups (ANOVA, P < 0.05) (Fig. 6). The difference in frequency of mEPSCs among the different groups can result from either a difference in density of synapses or a difference in transmitter release probability (as calcium stores may affect neurotransmitter release). We therefore compared the responses of 18 control and eight DN cells to pulse application of high osmolality medium (300 mm sucrose added to the normal recording medium) near the recorded neurons. In both groups, there was a marked increase in mEPSC frequency during the presence of high osmolality medium, although there was no difference in the magnitude of the difference between the groups, indicating that the difference does not reflect a presynaptic release probability (data not shown). These results indicate that the DN group of neurons that express immature spines are not forming as many synaptic connections as the control or CA groups and, thus, that Orai1 has an important role in maturation of dendritic spines and the formation of their functional connections.

Chemical LTP

We extended these observations to explore whether the DN transfected neurons are also less capable of undergoing morphological changes following induction of chemical LTP, as produced by NMDA‐enhancing conditioning medium. These experiments were conducted with only two groups for comparison, and we assumed that the CA and the Rescue groups are less important such a comparison. Cultures were imaged at 3D in standard medium, followed by a 3 min exposure to the conditioning medium, which was subsequently replaced by the standard medium, with the cells being imaged again 1 h later. Six control neurons (in three different cultures) expressed 6.5 ± 0.96 new mushroom spines (both short and long ones) per dendritic segments of 100 μm, and fewer (4.3 ± 1.4) new filopodia, as reported previously (Korkotian et al. 2014). By contrast, six DN neurons grew significantly fewer new spines (1 ± 0.44) but many more filopodia (15 ± 2.38) compared to controls. The difference was highly significant (Fig. 7 C). These results indicate that, although the DN cells are responsive to the change in ambient excitatory tone, they are unable to consolidate their response into the production of new mushroom‐type dendritic spines, as is the case for control cells.

Figure 7. Chemical LTP induces formation of new dendritic spines in control but not in DN transfected neurons.

Six control (A) and six neurons transfected with DN plasmid (B) were imaged before and 60 min after a 3 min exposure to conditioning medium which produces chemical LTP, reported in Korkotian et al. (2014). New spines and filopodia were counted in three fields of view of dendrites for each neuron. The scale bars in (A) and (B) are both 1 μm. C, a highly significant increase in mushroom spines was seen in the control but not in the DN neurons. The opposite was seen when counting new filopodia. Not counted are the stubby spines, which were not different between the groups. Significance levels: ^*** P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com]

siRNA for Orai1

Because the plasmids that have been used in the present study are added to the native molecules, it was imperative to demonstrate that a change in intrinsic Orai1 also produces the structural/functional changes observed above. For this purpose, we used the siRNA methodology to interfere with intrinsic Orai1. In these studies, siRNA targeted to Orai1 was compared with non‐targeted control RNA (Fig. 8). First, using western blots, it was confirmed that the siRNA‐treated neurons express a marked reduction in Orai1 (Fig. 8 A). None of the other proteins (Orai2, STIM1 and STIM2) was affected by the presence of siRNA to Orai1. This was validated using immunofluorescence staining for Orai1 (Fig. 8 B). We then resorted to electrophysiological recording from 21 non‐targeted and 25 targeted neurons to reveal a significant reduction in mEPSC frequency, and a tendency for smaller amplitude of mEPSCs (Fig. 8 C–E). There was no difference between non‐targeted and siRNA‐treated neurons in input resistance (185 ± 16 MW, n = 17 and 165 ± 11 MW, n = 20, in the two groups, respectively, P > 0.3). A Sholl analysis measured in perimeters of 40 and 80 μm around the somata of GFP transfected neurons that were co‐transfected either with the targeted or non‐targeted siRNA plasmids showed no significant difference between the dendritic arbors of the two groups of neurons (10 and 11 neurons: 8.2 ± 1.1 and 6.6 ± 0.7 processes in the 40 μm perimeter and 8.1 ± 1.022 and 6.6 ± 0.9 processes in the 80 μm perimeter, for the targeted and non‐targeted groups, respectively; images not shown, P > 0.05).

Figure 8. Exposure of neurons to siRNA targeted to Orai1 (T) or non‐targeted (NT) control RNA affects cell morphology and function.

A, western blot analysis of the effect of T on Orai and STIM peptides. Although T caused a marked and highly significant reduction in Orai1, it did not affect the density of Orai2, or the two STIMs. b‐actin was also the same in the two groups measured. B, fluorescence images of dendritic segments exposed to T and NT RNA, counterstained for Orai1 (red). Note that most of the spines in the NT cell are endowed with Orai1 puncta, unlike those exposed to the siRNA (T). C, sample records of mEPSCs taken from NT (left) and T (right) neurons. D and E, averaged amplitudes (D) and frequency (E) taken from 21 control (NT) and 25 siRNA (T) groups. F, density of spines per 100 μm dendritic segments in the T and NT groups. Spines were divided into mushroom (M), long mushroom (L) stubby+short (S) and filopodia (F) types. A clear difference between the T and NT groups in the M vs. F groups is obvious. G, analysis of immunohistochemical staining for Orai1 (as in B) shows a significant difference between the T and NT groups with respect to both the spines and parent dendrites (209 puncta from 16 cells each group). Significance levels in (E) to (G): ^* P < 0.05, ^** P < 0.01, ^*** P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com]

The morphology of dendritic spines was then compared between the two groups in 44 fields of targeted (T) and 59 fields of non‐targeted (NT) siRNA‐treated neurons. There was a highly significant difference between the two groups with respect to the proportion of mature mushroom spines and filopodia; although most of the protrusions in the T cells were of the immature, filopodia type, approximately one‐third of the protrusions in the cells exposed to the NT RNA were of the mature type, as seen in cells unexposed to external RNA (44 and 59 fields in the targeted and non‐targeted groups, respectively, P < 0.008), whereas there was no difference in the stubby/short spines (Fig. 8 F). As expected, the intensity of Orai1 staining in the spines and dendrites of the NT group was much higher than that in the T group (comparing 209 spines from 16 cells in each group, P < 0.001) (Fig. 8 G). These data confirm the role of Orai1 in dendritic spine formation and maturation.

Discussion

The results of the present study demonstrate a novel role for Orai1 in the regulation of dendritic spine maturation and plasticity. It has long been demonstrated that Orai1 is a membrane bound, store‐operated calcium channel, allowing calcium entry into the cell upon store depletion. A reduction in store calcium is sensed by the STIM calcium sensor, which assumes a punctate shape, moves from the endoplasmic reticulum to the membrane, links to Orai and activates it to allow influx of calcium ions. Our results indicate that both dendrites and spines are endowed with clusters of Orai1 molecules (= puncta), which respond to a transient surge of [Ca²⁺]_o by a rise in [Ca²⁺]_i, provided that the stores are emptied. The time course and magnitude of the effects, as well as the presence of Orai1 puncta in dendritic spines, indicate that the spine compartment may possess the machinery for influx of calcium via the Orai1 channel, independently of the dendritic shaft, and probably independently of voltage‐ and glutamate‐gated calcium influx. The amount of [Ca²⁺]_i raised in response to the surge of [Ca²⁺]_o is much smaller in Orai1(–) spines in the same neuron, as well as in DN‐Orai1 transfected neurons, indicating a role for Orai1 in the cellular response to the [Ca²⁺]_o surge. These results confirm previous observations on the role of STIM/Orai1 in SOCE in central neurons, which is now also extended to the dendritic spine compartment. Furthermore, our results indicate that Orai1 may have an important role in developing 10‐day‐old neurons (Fig. 5) and not in mature 20‐day‐old neurons where other calcium entry channels may have a more significant role in spine plasticity. At the later age, spines are already endowed with presynaptic terminals, confirming that, at this age, Orai1 may already be redundant as a major calcium entry channel.

It should be emphasized that, although we confirm the dependence of STIM/Orai complex on ambient [Ca²⁺]_o, we did not investigate the involvement of STIM in spine growth, maturation and plasticity, and the results of the present study therefore address only the role of Orai1 in these functions. Interestingly, although there is evidence for function of both STIM1 and STIM2 in hippocampal neurons (Zhang et al. 2014; Zhang et al. 2015), our co‐localization data indicate that STIM2 has a greater probabilty than STIM1 of moving into dendritic spines and teaming up with Orai1. This confirms earlier suggestions that STIM2 is more abundant in hippocampal neurons than STIM1 (Gruszczynska‐Biegala et al. 2011) and that it also may play a role in calcium regulation, in that it responds to smaller changes in endoplasmic [Ca²⁺] and is able to link and activate Orai1 (Brandman et al. 2007), resulting in a larger response to calcium depletion than STIM1 (Gruszczynska‐Biegala et al. 2011). Similarly, STIM2 may have an important role in regulation of dendritic spine morphology (Zhang et al. 2015). On the other hand, Orai2 has a lower probability of serving a significant role in dendritic spine structure and function, and its distribution in the neuron does not favour a spine location. Further studies are needed to examine the role of Orai2 in neuronal functions (Hoth and Niemeyer, 2013).

Surprisingly, in neurons transfected with dominant negative Orai1, spines do not mature, and remain as filopodia, and they do not produce as many functional synaptic connections as the normal controls or neurons transfected with the CA form of Orai1. Furthermore, the presence of Orai1 puncta can predict the local formation of a dendritic spine in response to a transient rise in ambient [Ca²⁺]_o. These results indicate that Orai1 may have a direct role in the formation and maturation of synaptic contacts. Intuitively, this channel should operate primarily under extreme conditions, when calcium stores are depleted (e.g. during epileptic seizure) (Steinbeck et al. 2011), and when normal synaptic activity is not sufficient to replenish the stores. This is probably not the case under normal conditions in mature neurons, where [Ca²⁺]_i is regulated to low levels through several mechanisms, including stores, calcium‐binding proteins, calcium pumps and transporters (Verkhratsky 2005) and calcium stores have a lower probability of being depleted. However, this can prove to be be important in immature neurons, having sparse synaptic connections, and lower levels of network activity.

The role of Orai1 in developmental regulation of synapse formation is especially surprising in view of the fact that dendritic spines can raise [Ca²⁺]_i to very high levels by influx of very few calcium ions through voltage‐ and ligand‐gated calcium channels. Thus, the importance of Orai1 in these functions is not altogether obvious. In a similarly small compartment, STIM/Orai1 complex is needed for growth cone motility (Mitchell et al. 2012), under conditions where no synaptic connections are playing any significant role, and the release of calcium from stores is assumed to be essential for growth cone advancement and the formation of connections. In our experiments, a lack of Orai1 in DN transfected or in siRNA‐treated neurons retards the development of dendritic spines, as well as the maturation of their synaptic contacts. Even at the short time scale, the presence of Orai1 puncta facilitates the formation of novel spines following a conditioning protocol, and can even predict where spines will form on the dendrite. This is related to the general issue of what triggers the formation of a spine in specific locations along the dendritic shaft. It is proposed that the presence of an Orai1 puncta can trigger local influx of calcium, which is sufficient to activate the molecular cascade, including actin polymerization, the formation of the postsynaptic density and the movement of glutamate receptors, which will then allow further influx of calcium through voltage‐ and NMDA‐gated channels, thus stabilizing the dendritic spine and forming connections with presynaptic terminals. In line with this assumption, spines or dendritic segments that contain Orai1 generate a larger response to a surge of [Ca²⁺]_o than spines that do not contain Orai1, or for which the Orai1 is of the DN species. These data indicate that Orai1 channels have a primary role in formation of synapses, and propose that voltage‐ and ligand‐gated channels play a role only later in the process of synapse formation and plasticity. This is consistent with the suggestion that Orai1 is important for growth cone formation, when synapses are not yet effective in growth cone motility.

A similar role to that of Orai1 in synaptic plasticity was ascribed previously by us and others to synaptopodin, a protein found in the base of dendritic spines and that is also assumed to constitute a component of endoplasmic reticulum calcium stores (Korkotian et al. 2014). Indeed, we found that Orai1 and synaptopodin co‐localize in some dendritic spines, indicating that they may share a role in regulation of stored calcium. Further experiments are needed to examine this possibility.

The possible role of SOCE channels in neuronal plasticity was proposed by Baba et al. (2003). Although Orai1 was not yet identified as a molecule with a role in plasticity, it was reported that a rise in dendritic [Ca²⁺]i is suppressed by SOCE antagonists, and that LTP is significantly reduced by these agents. Thus, Orai1 may play a role in neuronal plasticity not only in developing neurons, but also in the mature brain.

On the other extreme is the potential role of STIM/Orai in the facilitation of destructive processes associated with ischaemia (Zhang et al. 2014), oxidative stress (Henke et al. 2013) or epileptic seizures (Steinbeck et al. 2011). In these cases, blockade of the Orai1 channel can protect cells from the destructive process associated with an uncontrolled rise in [Ca²⁺]_i following traumatic neuronal events. This may be related to glutamate toxicity (i.e. neuronal death following a glutamate‐induced rise of [Ca²⁺]_i), which is facilitated by Orai, and reduced in neurons where Orai is downregulated (Henke et al. 2013). This may underlie the detrimental effects of excess glutamate that lead to dendritic spine collapse (Schubert et al. 2006; Sala and Segal, 2014). Thus, Orai channels constitute a double‐edged sword, demonstrating constructive and destructive actions, depending on the state of the neuron and its calcium load. The dendritic spine is assumed to play a central role in this regulation.

Additional information

Competing interest

The authors declare that they have no competing interests.

Author contributions

EK and MS designed the experiments, conducted the experiments, analysed the data and wrote the paper. EB prepared the cultures, western blots, plasmids and siRNAs. All authors have read and approved the final copy and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

The present study was supported by a grant from the Israel Science Foundation.