Store-operated Ca2+ entry is activated by every action potential in skeletal muscle

Store-operated calcium (Ca2+) entry (SOCE) in skeletal muscle is rapidly activated across the tubular system during direct activation of Ca2+ release. The tubular system is the invagination of the plasma membrane that forms junctions with the sarcoplasmic reticulum (SR) where STIM1, Orai1 and ryanodine receptors are found. The physiological activation of SOCE in muscle is not defined, thus clouding its physiological role. Here we show that the magnitude of a phasic tubular system Ca2+ influx is dependent on SR Ca2+ depletion magnitude, and define this as SOCE. Consistent with SOCE, the influx was resistant to nifedipine and BayK8644, and silenced by inhibition of SR Ca2+ release during excitation. The SOCE transient was shaped by action potential frequency and SR Ca2+ pump activity. Our results show that SOCE in skeletal muscle acts as an immediate counter-flux to Ca2+ loss across the tubular system during excitation-contraction coupling.

S tore-operated Ca 2+ entry (SOCE) is a mechanism that is slowly activating in most cells, where stromal interacting molecule 1 (STIM1) molecules sense Ca 2+ depletion in the internal Ca 2+ store and migrate to plasma membrane contact sites to trigger Ca 2+ influx through Orai1 channels 1 . In contrast, SOCE has evolved to become a fast mechanism in skeletal muscle, probably in parallel with the evolution of the highly organised membrane structures that support the rapid transients of Ca 2+ in the fibre evoked by action potentials 2 . In muscle, the surface plasma membrane of the fibre has regular invaginations, to form the tubular (t-) system. The dominant element of the t-system, the transverse tubules reach each sarcomere to form junctions with the sarcoplasmic reticulum (SR) terminal cisternae 3 where the transverse tubular dihydropyridine receptor (DHPR)/L-type Ca 2+ channel directly activates the SR ryanodine receptor (RyR) to rapidly activate Ca 2+ release 4 . This process is known as excitation-contraction coupling (EC coupling) and provides the basis for muscle contraction.
While the presence of a robust and fast SOCE in muscle is clear, the physiological activation and primary roles of SOCE in this tissue remain elusive 2,5 . The biophysical properties of SOCE in muscle 2 suggest its physiological role can be an immediate one, related to compensatory Ca 2+ movements during action potential-evoked Ca 2+ release. Knowledge of the physiological patterns of store-dependent influx and the conditions associated with its activation are critical in defining physiological role, which will help in our understanding of SOCE-related myopathies, developmental defects and fibre-type shifts in STIM1 and ORAI1 knockout mice [6][7][8][9][10] .
Ca 2+ within the SR ([Ca 2+ ] SR ) is strongly buffered by calsequestrin and the high-capacity SR Ca 2+ pumps 11 . This prevents steady state or tonic activation of SOCE in resting fibres 5,12 . Furthermore, the SR strongly buffers Ca 2+ during EC coupling, preventing deep depletion of the [Ca 2+ ] SR 13-15 , making it difficult to understand how SOCE is activated physiologically in skeletal muscle, if we continue to assume that a bulk depletion of [Ca 2+ ] SR is required for SOCE activation 5 .
We hypothesized that SOCE could be activated during action potential-evoked Ca 2+ release by local depletion of Ca 2+ immediately behind the RyRs at the SR terminal cisternae. Local depletions of Ca 2+ inside the SR would briefly and uniformly dissociate Ca 2+ from the local population of STIM1 7,16 to rapidly active SOCE. The rapid transduction of such a signal requires the static junction between the transverse tubules and the SR terminal cisternae membranes, allowing RyR activity to intimately link EC coupling and SOCE 2 . Consistent with this, it has been demonstrated that only the transverse tubules support SOCE; whereas, the longitudinal elements of the t-system that lack proximity to RyRs have become dynamic Ca 2+ -buffers, devoid of SOCE 17 .
Here we use a microscopy-based technique with the necessary sensitivity and temporal resolution to measure the occurring Ca 2+ -fluxes across the t-system membrane 18,19 under physiological stimulation patterns. This approach allowed us to identify that the physiological activation of SOCE is by every action potential-induced Ca 2+ release, presenting as a phasic Ca 2+ influx. This finding allows the elucidation of the physiological roles of SOCE in muscle and broadens our understanding of the properties of STIM1 and SOCE more generally.

Results
t-system membrane Ca 2+ fluxes following every action potential. To resolve Ca 2+ fluxes across the t-system membrane during action potential-evoked Ca 2+ release required imaging Ca 2+ in the cytoplasm and t-system with high sensitivity and millisecond temporal resolution. To do this we used mechanically skinned fibres (Fig. 1a, Methods section), in which we trapped the Ca 2+ -sensitive fluorescent dye rhod-5N in the t-system of skinned fast twitch extensor digitorum longus (EDL) fibres from the rat, as previously described 19 . This preparation preserves intact EC coupling 13,20 and respective Ca 2+ release and force responses are indistinguishable from those in intact fibres 19,[21][22][23] . The removal of the sarcolemma during the skinning process grants access to the cytosol and allows for the measurement of the intracellular Ca 2+ levels with the use of a spectrally separate Ca 2+ -sensitive dye from rhod-5N, fluo-4. Figure 1b is an example of spatially averaged profiles of cytoplasmic fluo-4 and t-system-trapped rhod-5N fluorescence during electrical field stimulation of a skinned fibre at 1, 2 and 5 Hz using the approach described in Fig. 1a. In each of these cases, field stimulation induced a transient rise in cytosolic Ca 2+ (Fig. 1b, grey trace). The simultaneous recording of rhod-5N fluorescence showed a decline, indicating a concurrent influx of Ca 2+ from the t-system to the cytoplasm with each action potential (Fig. 1b, black trace). Both signals returned to their respective baseline levels following the cessation of stimulation. Note that the cytoplasmic solution contained 10 mM EGTA with free [Ca 2+ ] cyto buffered to 50 nM. The introduction of EGTA as the dominant cytoplasmic Ca 2+ -buffer allows the calculation of the amount of total Ca 2+ released from the SR (Methods).
Close inspection of the [Ca 2+ ] t-sys signals shows a fast decline in t-system Ca 2+ associated with each action potential-evoked Ca 2+ release (Fig. 1b). Termination of Ca 2+ efflux from the t-system allows the signal to recover toward baseline in between individual stimulations pulses, generating a saw-tooth like pattern. As the time for recovery of [Ca 2+ ] t-sys became too short at increasing stimulation frequencies, there was a progressive summation of individual [Ca 2+ ] t-sys depletions to reduce [Ca 2+ ] t-sys in a frequency-dependent manner (Fig. 1b). The overall decline in the [Ca 2+ ] t-sys signal during repetitive electrical stimulation could be fitted by an exponential function. A summary of the rate constants at the stimulation frequencies applied is displayed in Fig. 1c. A positive relationship between stimulation frequency and the rate constant of [Ca 2+ ] t-sys depletion was observed. At stimulation frequencies of 20 and 50 Hz, the temporal resolution of the recordings (18 ms, Methods section) omitted to resolve individual [Ca 2+ ] t-sys depletion events. Still the depletion of [Ca 2+ ] t-sys could be fitted by exponential functions (Supplementary Figure 1). The coherence to the observed relationship is consistent with a successful electrical stimulation at these high frequencies and indicates a coupled Ca 2+ influx from the t-system to the cytoplasm with each action potential under these conditions.
When we evaluated the [Ca 2+ ] t-sys depletion, as well as the recovery of the individual steps for the tested frequencies we noticed that individual depletion steps were getting smaller with each applied pulse (Supplementary Figure 2a). We reasoned that this would be due to the decreasing levels of [Ca 2+ ] t-sys , the driving force for Ca 2+ entry 24 Figure 2b). Thus, the depletion steps and with it the underlying calcium flux were directly proportional to the Ca 2+ gradient across the t-system (neglecting the much smaller cytosolic Ca 2+ concentration) suggesting that the observed Ca 2+ flux was carried by a Ca 2+ -selective ion channel 24 .
In order to improve the temporal resolution of our signal, we performed linescans, accepting a dramatic loss in signal to noise ratio due to the loss of the spatial averaging. To partly restore for the deteriorated signal we averaged at least 10 consecutive linescans aligned by the peak of the action potential-induced Ca 2+ release. This provided an improvement in temporal resolution to 80 μs. With this approach, we determined the activation of the Ca 2+ influx during a single-action potential occurred with a time constant of 10 ms (Supplementary Figure 2c and 2d).
Every action potential activates SOCE. The known t-system Ca 2+ fluxes that could be carrying the observed phasic influx during EC coupling are L-type Ca 2+ channel or SOCE. To determine whether the phasic Ca 2+ flux was carried by the L-type Ca 2+ channel or not we applied the known antagonist nifedipine and the agonist BayK8644 and assessed [Ca 2+ ] t-sys depletion during trains of action potentials under these conditions. Neither nifedipine, at a close to saturating concentration (10 μM) 25,26 , nor BayK8644 had any effect on the observed phasic [Ca 2+ ] t-sys depletion nor its rate of depletion (Fig. 2). These results exclude the possibility that the L-type Ca 2+ channel is responsible for carrying the observed phasic t-system Ca 2+ flux.
Next, to test whether the phasic influx was store-dependent, we examined whether interfering with store-depletion during excitation of the fibre affected the phasic influx. To do this we blocked the RyRs directly, to prevent Ca 2+ release during field stimulation, using three independent RyR antagonists: tetracaine, ryanodine, and raised free [Mg 2+ ] cyto . This approach was expected to separate t-system Ca 2+ fluxes dependent directly on excitability or Ca 2+ release. First, we tried tetracaine. As expected, 10 μM of the RyR inhibitor tetracaine inhibited the electrically evoked cytosolic Ca 2+ transients. This concurrently reduced the amount of Ca 2+ depleted from the t-system (Fig. 3). Raising the tetracaine concentration to 30 μM abolished the cytosolic Ca 2+ transients and the phasic depletions of [Ca 2+ ] t-sys (Fig. 3). Interestingly, although individual depletion steps could not be resolved anymore under these conditions, it was evident that [Ca 2+ ] t-sys steady-state values were different before and after electrical stimulation. Thus, electrical stimulation was able to deplete some [Ca 2+ ] t-sys even in the absence of any detectable SR Ca 2+ release. This result is consistent with maintained excitability of the t-system following action potentials in the presence of μM  trapped in the t-system and loaded into the cytosol, respectively, were spatially averaged and calibrated (Methods section). Electrical stimulation at 1, 2 and 5 Hz as indicated was started shortly after beginning of the recording and was continued until a new steady state in the t-system was reached, unambiguously identifiable by the termination of cytosolic transients. Note that the employed sampling rate of 55 frames s −1 is below the Nyquist criterion, which results in an apparent modulation of cytosolic Ca 2+ transient amplitudes. Green lines represent mono-exponential fits to the decaying phase of [Ca 2+ ] t-sys upon stimulation. c A summary of the respective rate constants as derived from the exponential fits for all tested frequencies. Data are derived from 10 fibres and given as mean ± SEM. EDL, extensor digitorum longus tetracaine and the activation of an apparently tonic, lowamplitude Ca 2+ current activated by the train of action potentials.
In the next experiments, raised cytosolic [Mg 2+ ] cyto and ryanodine were used to inhibit the RyR, both of which are known not to affect t-system excitability 27  We are aware that it is possible for ryanodine and tetracaine to affect the DHPR and voltage-gated Na + channel, respectively. However, the concentrations of ryanodine we have used is wellbelow the Ki for binding to the DHPR (45 μM) 28 . The presence of 10 μM tetracaine allowed action potential-induced Ca 2+ release to remain active, though reduced. Furthermore, the use of raised Mg 2+ as a RyR antagonist is known not to affect t-system excitability 29,30 . Regardless, the presence of the antagonist did not affect the maintenance of excitability as we observed the tonic Ca 2+ influx in each case, which likely reflects L-type Ca 2+ channel influx 31 .
Additionally, we examined the possibility that the Ca 2+ released from the SR into the junctional space per se rather than SR Ca 2+ depletion caused activation of phasic t-system Ca 2+ flux. To this end we manipulated the [Ca 2+ ] in the junctional space during SR Ca 2+ release by bathing the fibre in a solution containing 10 mM BAPTA instead of EGTA. BAPTA provides a much faster rate of Ca 2+ chelation. Under these conditions we observed a suppression of the cytosolic Ca 2+ transients during EC coupling, as expected, but also observed the persistence of the phasic Ca 2+ influx from the t-system (Supplementary Figure 5). The rate constant from exponential curves fitted to the decline in [Ca 2+ ] t-sys during 2 Hz stimulation and the plateau of [Ca 2+ ] t-sys depletion were not significantly different from that in EGTA (two-tailed t-tests, p = 0.4728 and 0.5833, respectively). This is consistent with depletion of Ca 2+ from within the SR as the trigger for the phasic t-system Ca 2+ flux and strongly argues against a mechanism of Ca 2+ -induced Ca 2+ influx, that could have potentially been carried by members of the TRP family of ion channels Together, these experiments (Figs. 2, 3 and Supplementary Figures 3-5) clearly demonstrate that the observed rapid [Ca 2+ ] t-sys depletion is dependent on the Ca 2+ depleted from within the SR.
To be consistent with the phasic flux being SOCE, the degree of SR depletion should be represented in the magnitude of the t-system Ca 2+ flux 2 . To determine the relationship between the amount of Ca 2+ released from the SR and the amount of Ca 2+ depleted from the t-system, we determined the total Ca 2+ released from the SR under control conditions and under conditions where Ca 2+ release was either suppressed or facilitated (for example, when Ca 2+ release was suppressed in tetracaine, Fig. 3, see Methods section for calculations). A summary of this analysis is plotted in Fig. 4. A two component model fitted the data. For very small and non-detectable Ca 2+ release from the SR (left branch of the graph) there was a constant amount of about 0.14 mM Ca 2+ depleted from the t-system (with respect to t-system volume). When SR Ca 2+ release surpassed a threshold of 1.2 mM Ca 2+ (with respect to SR volume) depletion of the t-system Ca 2+ increased with increasing Ca 2+ released from the SR. A maximum plateau was reached at a value of about 0.3 mM (right branch of the graph). The amount of extracellular Ca 2+ entering the cytoplasm from the t-system was only a small fraction relative to the contribution of the SR during EC coupling. Ca 2+ released from the SR contributed about 99% to the Ca 2+ entering the cytoplasm (blue trace, right axis, Fig. 4). Taken together, this data show the presence of a store-independent mechanism (the apparently tonic flux) that can be observed in isolation when no or very little Ca 2+ was released from the SR and the emergence of a second, store-dependent mechanism (the phasic flux) when a certain threshold of SR Ca 2+ release was exceeded.     (Fig. 1) could no longer be resolved (Fig. 6). In the same fibre, the re-introduction of 10 mM EGTA allowed these fast [Ca 2+ ] t-sys depletions to be observed again (Supplementary Figure 6). Under 2 Hz field stimulation, the steady-state depletion of [Ca 2+ ] t-sys was only significant in high [EGTA] (Fig. 6b)

Discussion
The physiological activation mechanism and a primary physiological role of SOCE in skeletal muscle have not been defined, even though the molecular identities and biophysical properties of SOCE have been described in some detail 2,7-9 . The main obstacle in determining the physiological activation and role of SOCE in muscle has been the lack of direct measurements of SOCE during EC coupling under physiological conditions, which we have addressed (Fig. 1). Using this approach we were able to observe a phasic influx of Ca 2+ from the t-system lumen with every action potential at all stimulation frequencies tested here. The pathway responsible for the Ca 2+ influx was established as predominantly store-dependent channels by showing (i) a dependence on the amount of Ca 2+ depleted from the SR (Fig. 4), (ii) resistance to the L-type Ca 2+ channel antagonist and agonist, nifedipine and BayK8644 (Fig. 2), (iii) an elimination by the blockade of SR Ca 2+ depletion with three independent RyR antagonists during excitation ( Fig. 3 and Supplementary Figures 3, 4), (iv) fast kinetics consistent with a high off-rate of Ca 2+ for STIM1 (and inconsistent with the slow activation of the Ltype Ca 2+ channel) (Fig. 1), (v) a linear dependence on the tsystem Ca 2+ gradient (Supplementary Figure 2b), and (vi) persistence in the presence of either cytoplasmic EGTA or BAPTA (Supplementary Figure 5). In the absence of Ca 2+ release, repetitive field stimulation did allow the isolation of an additional mode of slowly activating, apparently tonic Ca 2+ influx during repetitive action potentials consistent with L-type Ca 2+ channel activation 31 (Fig. 3). Changing the Ca 2+ -buffering capacity of the cytoplasm also allowed the manipulation of the ability of the tsystem to extrude Ca 2+ during SR Ca 2+ release 19,32 . This exposed a role of rapidly activated SOCE during EC coupling was to act as an immediate counter-flux to t-system Ca 2+ extrusion to prevent the loss of fibre Ca 2+ during muscle stimulation. It has been implied from indirect evidence that SOCE is active during the declining phase of the Ca 2+ transient of a tetanus 9,15 . However, by directly measuring SOCE during EC coupling we show that this is not the case (Figs. 1-3). The physiological waveform of SOCE influx is phasic, following each action potential. The heavily Ca 2+ -buffered SR rapidly inactivated SOCE (Figs. 1-3). We could only observe a tonic activation of SOCE under non-physiological conditions of a poisoned SR Ca 2+ pump and hence depleted SR Ca 2+ stores (Fig. 5). During periods of fibre quiescence the electrochemical gradient for Ca 2+ across the t-system is sufficient for the passive leak of extracellular Ca 2+ into the fibre to set the required level of Ca 2+ inside the SR 5,33 . This is consistent with STIM1 or ORAI1 null muscle responding normally to field stimulation at 50 Hz, which implies a normal load of SR Ca 2+5,7,9,34 . Furthermore, during prolonged periods of muscle stimulation where the muscle experiences metabolic fatigue, the total amount of Ca 2+ inside the SR is still significant and is not the limiting factor in Ca 2+ release [35][36][37] .
The rapid activation of SOCE in conjunction with action potential-induced Ca 2+ release means only a relatively small quantum of the total Ca 2+ held within the SR gets released as the trigger SR Ca 2+ depletion for Ca 2+ dissociation from STIM1. The steady level of bulk [Ca 2+ ] SR does not drop significantly [13][14][15] and cannot be critical to SOCE activation. It must be the Ca 2+ dropping during release in a locally restricted domain behind the RyR channel pore that is critical for SOCE activation. Thus, we suggest that SOCE is rapidly activated by local depletions of Ca 2+ at the near membrane inside the SR terminal cisternae following the action potential-coordinated Ca 2+ release. This is, of course, consistent with a k off Ca 2+ from STIM1 of probably >2000 s −138 . Near membrane Ca 2+ depletion will uniformly occur at all the terminal cisternae of the fibre, where STIM1 and RyRs are likely in close quarters 2,16,24 . Under this condition the bulk [Ca 2+ ] SR during twitches and tetani can remain high [13][14][15] while SOCE becomes briefly active (Figs. 1-4). Our findings provide further support to the proposal that Orai1 and STIM1 are in prearranged complexes that can be rapidly activated upon Ca 2+ release 2,16,24,32 . Others have challenged this idea based on the increase of a lowbiomolecular fluorescence complementation (BiFC) signal in a resting fibre upon poisoning of the SR Ca 2+ pump and thorough depletion of SR Ca 2+9 . The conclusion was based on the assumption that only the stronger BiFC signal represents prearranged complexes. It is, however, possible that both the full and depleted SR have prearranged Orai1-STIM1 complexes, differing only in absolute number, and therefore BiFC signal.
Our line of argument of pre-coupled Orai1-STIM1 complexes is further supported by the observation that SOCE activated with a time constant of only 10 ms (Supplementary Figure 2c, d). In fact this makes it hard to envision STIM1 and Orai1 physically separated at the junctional membranes, given that respective activation kinetics of uncoupled Orai1-STIM1, as has been demonstrated in multiple non-muscle tissues, lie in the seconds range 1 .  The simultaneous imaging of Ca 2+ in the cytoplasm and SR lumen has shown that the nadir of Ca 2+ depletion inside the SR lags behind the closure of RyRs during sparks and action potential-induced Ca 2+ release by 50 and 30 ms, respectively 13 . To explain the anomaly of SR Ca 2+ depletion continuing after the cessation of Ca 2+ release, the source of the Ca 2+ during these events was proposed to originate from a locally restricted domain near the membrane behind the RyR channel pore whose diffusional properties differ from the bulk SR 13 . As such, the observed delay of the nadir of the measured bulk SR Ca 2+ depletion must be associated with the equilibration of these local Ca 2+ gradients. Given the length of the delay between closure of the RyRs and the time to [Ca 2+ ] SR nadir, it is reasonable to suggest that there is a barrier to Ca 2+ diffusion from the bulk SR to the near membrane as Ca 2+ is released. This is likely to be calsequestrin (CSQ), given the positioning of CSQ at the SR terminal cisternae, immediately behind the RyRs and the unique Ca 2+ -adsorbing and high Ca 2+ -buffering capacity of CSQ [39][40][41] . Thus the spatiotemporal separation of Ca 2+ sources inside the SR makes the near membrane a privileged signalling environment where Ca 2+ can rapidly and locally deplete, with the opening and closing of RyRs. These transient depletions of Ca 2+ near the membrane can happen without major depletions of the bulk [Ca 2+ ] SR be it during a spark, twitch or punctuated throughout a tetanus [13][14][15] . This system provides the environment for the rapid dissociation of Ca 2+ from STIM1 in the near membrane 16 for activation of SOCE following every action potential, as we have demonstrated (Figs. 1-4). Consistent with this, preliminary reports using Ca 2+ indicators targeted close to the RyR channel within the SR terminal cisternae lumen have revealed that local Ca 2+ release is greater and faster than that occurring in the bulk SR lumen during EC coupling 42 .
In respect thereof, the abundance and positioning of CSQ, RyRs and STIM1 in this environment are paramount to muscle Ca 2+ homoeostasis and, ultimately, muscle physiology and health (e.g., refs. 6,22,[43][44][45]. While the normal positioning of CSQ and RyRs has been shown 39,46 , the exact positioning of STIM1 is unknown because of the lack of a specific STIM1 antibody to use for localization imaging 32 . Developing a picture of how STIM1 and RyRs position themselves relative to each other at the terminal SR and identifying how this arrangement changes under different conditions will be required to gain a further understanding of signalling at this nano-domain. However, our highspatiotemporal imaging of Ca 2+ transients (Figs. 1-4) during physiological Ca 2+ release limits the possible location of the active proteins in the fibre to the junctional membranes, nearby the RyRs.
We propose that a primary physiological role of SOCE is an immediate, rapidly activated counter-flux to t-system Ca 2+ extrusion in skeletal muscle fibres that is activated in response to the local depletion of SR terminal cisternae Ca 2+ as evoked by action potentials. In the presence of a weakly Ca 2+ -buffered cytoplasmic environment, there was only a small net change in the [Ca 2+ ] t-sys during a train of action potentials (Fig. 6). This level of Ca 2+ -buffering is equivalent to the situation in the muscle, so allowing a build-up of Ca 2+ at the cytoplasmic side of the t-system membrane during EC coupling to cause activation of Ca 2+ transport from the junctional space to the t-system lumen. The simultaneous activation of SOCE with t-system Ca 2+ extrusion represents an efficient mechanism to keep Ca 2+ in the fibre during any period of stimulation. The advantage of the phasic Ca 2+ influx being store-dependent is that the recovery of Ca 2+ that enters the t-system during EC coupling is tightly regulated by the amount of Ca 2+ exiting the SR, allowing balance of fluxes, or net Ca 2+ influx that may assist muscle activity. Tracking the entry of radioactive Ca 2+ into intact muscle during continuous, low-frequency stimulation has shown an accumulation of Ca 2+ in the muscle 47 , suggesting that Ca 2+ influx is greater than efflux in vivo 48 . The major component of this Ca 2+ influx would be SOCE at the transverse tubules in conjunction with excitation (Figs. 1-4) 47 . The net movement of Ca 2+ across the t-system is probably tuneable by regulation of the relative activity of the Na + -Ca 2+ exchanger, plasma membrane CaAT-Pase and SOCE. It remains possible that deviation from the Ca 2+ homoeostasis we describe here may alter signalling cascades 10 .
Our demonstration that the normal activation of SOCE is immediate following every action potential-evoked release of Ca 2+ in healthy muscle provides the framework in which to understand SOCE-related myopathies. e.g., ref. 6 . It is not currently clear how quickly the absence of SOCE as a counter-flux will compromise fibre Ca 2+ levels during bouts of EC coupling, as these fluxes are small (Fig. 4). Certainly, it is likely that changes in Ca 2+ levels in the nano-domains of the junctional membranes during muscle activity will be sensitive to SOCE. The in-phase nature of the SOCE influx with the pattern of action potential stimulation (Figs. 1-4) also makes it possible that the SOCE transients entering the junctional space are an essential part of phasic Ca 2+ signalling in response to muscle usage patterns [8][9][10]49,50 , as local Ca 2+ signals activate diffusible messengers 51 .
Taken together, our results suggest the primary physiological role for SOCE in skeletal muscle to be a counter-flux to t-system Ca 2+ extrusion, to maintain Ca 2+ homoeostasis. Our observation of the control of t-system Ca 2+ -handling by action potentialevoked Ca 2+ release can explain the classic observation of Armstrong et al. 52 , where skeletal muscle twitched for long periods in the presence of EGTA and virtual absence of external Ca 2+ . Under these conditions some of the Ca 2+ released from SR is extruded across the t-system membrane to the t-system lumen. Ca 2+ entering the t-system lumen requires seconds to diffuse from the fibre 53 . The simultaneous opening of the storedependent channels with SR Ca 2+ release (Figs. 1-4) turn the t-system into a sieve, causing the Ca 2+ that entered the t-system lumen during Ca 2+ release to re-enter the cytoplasm. That is, SOCE captures Ca 2+ that is extruded to the t-system and returns it to the cytoplasm before it would otherwise be lost to the external environment beyond the t-system lumen.

Materials and methods
Preparation of skinned fibres and solutions for Ca 2+ imaging. All experiments on rats were approved by the Animal Ethics Committee at The University of Queensland. Male Wistar and Sprague rats 4-8 months of age (University of Queensland Biological Resources, Brisbane) were killed by asphyxiation in CO 2 . The extensor digitorum longus (EDL) muscle was rapidly excised and placed in a Petri dish under paraffin oil above a layer of Sylgard. Small bundles of fibres were exposed to a Ringer solution containing 2.5 mM of the Ca 2+ -sensitive dye rhod-5N for at least 10 min. Thereafter, segments of individual fibres were isolated and the sarcolemma was mechanically removed with forceps (fibre skinning). The skinned fibres were mounted in a custom built chamber on top of a 1. The experimental chamber was mounted above a ×20 objective of a Leica TCS SP8 laser scanning confocal microscope system equipped with a 12 kHz resonant scanner. Rhod-5N and Fluo-4 were excited by 561 and 488 nm laser lines, respectively, and emitted light was collected with high-sensitivity GaAsP detectors in bands of 40 nm. xyt image series of the two colours were taken in sequential mode with an individual xy image size of 512 × 128 pixels and a time resolution of 18 ms (55 frames s −1 ). Pinhole was set to 1.5 AU. Fluorescence intensity across the imaged fibre area was averaged for both colours to derive traces of the mean fluorescence intensity over time that were then converted to [ F min was determined at the end of each experiment by exposing the fibre to a calcium free solution. F max was calculated from the baseline fluorescence in the standard solution containing 50 nM of free [Ca 2+ ] by rearranging equation (1): Determination of the amount of total Ca 2+ released from the SR during an electrically evoked action potential. This estimation is based on a method originally described by Melzer et al. 55 . Briefly, the cytosolic Ca 2+ signals in 10 mM EGTA show a sharp transient followed by a slow decay upon each action potential (Fig. 1). The sharp transient is caused by the release of Ca 2+ from the SR, which is incompletely buffered by EGTA while the slow decline reflects the diffusion of free Ca 2+ out of the imaged region of interest, i.e., from within the fibre. Assuming the diffusive loss during the short time of an action potential to be negligible the change in cytosolic free Ca 2+ levels before and right after an action potential directly reflects the additional amount of Ca 2+ released from the SR. Considering a simple Ca 2+ -EGTA buffer system (Ca 2+ +EGTA<->Ca 2+ EGTA) and neglecting the much smaller endogenous buffering capacity 56 the total amount of Ca 2+ present can be calculated from the free [Ca 2+ ] cyto according to the following equation: And the amount of total calcium released by the SR is obtained by: where [Ca 2+ ] before and [Ca 2+ ] after are the free calcium concentrations right before and right after each action potential, K D is the dissociation constant of EGTA (200 nM) 56 , and [EGTA] T is the total EGTA buffer concentration. The so converted values refer to the cytosolic volume of the fibre and were converted to values referring to the SR volume by assuming a cytosol to fibre volume ratio of 0.7 and a SR to fibre volume ratio of 0.093 57 .
Calibration of t-system rhod-5N fluorescence. Calibration of rhod-5N fluorescence was performed as previously reported 19 . Fluorescence values were converted to [Ca 2+ ] t-sys using equation (1) with a K D of 0.872 mM 19 .
Determination of t-system depletion and recovery values.  [2][3], [3][4], [4-oo] mM. Data were fit with a nonlinear least square routine to a segmented equation consisting of an initial constant baseline followed by an exponential rise according to the formula: with x 0 indicating the start of exponential rise.
Back calculation of Ca 2+ movements across the t-system membrane. Each action potential was accompanied by a depletion step, as well as a reuptake step. Upon repetitive stimulation [Ca 2+ ] t-sys eventually reached a new steady state, a condition in which the single depletion and reuptake steps must exactly balance each other (Supplementary Figure 2a). This relationship can be used to obtain the amplitude of the first depletion step under conditions in which the individual steps cannot be resolved anymore. The change in t-system steady-state Ca 2+ level (Δ[Ca 2+ ] SS t-sys ) during repetitive stimulation is the sum of all individual depletion and re-uptake steps, i.e.
Δ Ca 2þ Â Ã SS tÀsys ¼ with Yd i and Yr i being the i th amplitude of depletion and re-uptake, respectively. The amplitudes of the individual depletions steps followed an exponential, while the amplitudes of the reuptake remained constant (Supplementary Figure 2a), thus Summing of the geometric series as obtained by inserting expression (7) into (6) yields an expression for the first depletion step in a train of stimulation as a function of Δ[Ca 2+ ] SS t-sys : Super events. On a few occasions electrical stimulation did not result in the generation of only a single action potential but was accompanied by a second action potential or even a small burst like discharge. Also, spontaneous action potentials occasionally seen during experiments happened to sometimes coincide with the electrical stimulation. Although these super events released considerably larger amounts of Ca 2+ into the cytosol as determined by the fluo-4 signal the respective depletion of the t-system was only moderately increased reaching a limiting level of about 0.3 mM.
Statistical analysis. For comparing two groups of data a two-tailed Student's t-test was used. For comparing more than two groups ANOVA with Tukey's post hoc test was used. Paired testing was applied when the data points to compare were sampled from the same fibres. All data are given as mean ± SEM.
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its supplementary information files.