Calsequestrin depolymerizes when calcium is depleted in the sarcoplasmic reticulum of working muscle

Significance We show that calsequestrin, the main Ca2+ storing protein of muscle, is polymerized inside the sarcoplasmic reticulum (SR) and its mobility increases greatly upon SR depletion, indicating depolymerization. Deep depletion causes massive calsequestrin migration and radical SR remodeling, often accompanied by a surge in intra-SR free Ca2+. The changes in calsequestrin polymerization observed in aqueous solutions therefore also occur in vivo. These changes help explain some uniquely advantageous properties of the SR as a source of calcium for contractile activation. The results support untested hypotheses about additional calsequestrin roles in the control of channel gating and facilitation of calcium flux. They also provide insights on the consequences of calsequestrin mutations for functional competence and structural stability of skeletal and cardiac muscle. Calsequestrin, the only known protein with cyclical storage and supply of calcium as main role, is proposed to have other functions, which remain unproven. Voluntary movement and the heart beat require this calcium flow to be massive and fast. How does calsequestrin do it? To bind large amounts of calcium in vitro, calsequestrin must polymerize and then depolymerize to release it. Does this rule apply inside the sarcoplasmic reticulum (SR) of a working cell? We answered using fluorescently tagged calsequestrin expressed in muscles of mice. By FRAP and imaging we monitored mobility of calsequestrin as [Ca2+] in the SR--measured with a calsequestrin-fused biosensor--was lowered. We found that calsequestrin is polymerized within the SR at rest and that it depolymerized as [Ca2+] went down: fully when calcium depletion was maximal (a condition achieved with an SR calcium channel opening drug) and partially when depletion was limited (a condition imposed by fatiguing stimulation, long-lasting depolarization, or low drug concentrations). With fluorescence and electron microscopic imaging we demonstrated massive movements of calsequestrin accompanied by drastic morphological SR changes in fully depleted cells. When cells were partially depleted no remodeling was found. The present results support the proposed role of calsequestrin in termination of calcium release by conformationally inducing closure of SR channels. A channel closing switch operated by calsequestrin depolymerization will limit depletion, thereby preventing full disassembly of the polymeric calsequestrin network and catastrophic structural changes in the SR.

Calsequestrin, the only known protein with cyclical storage and supply of calcium as main role, is proposed to have other functions, which remain unproven. Voluntary movement and the heart beat require this calcium flow to be massive and fast. How does calsequestrin do it? To bind large amounts of calcium in vitro, calsequestrin must polymerize and then depolymerize to release it. Does this rule apply inside the sarcoplasmic reticulum (SR) of a working cell? We answered using fluorescently tagged calsequestrin expressed in muscles of mice. By FRAP and imaging we monitored mobility of calsequestrin as [Ca 2+ ] in the SR-measured with a calsequestrin-fused biosensor-was lowered. We found that calsequestrin is polymerized within the SR at rest and that it depolymerized as [Ca 2+ ] went down: fully when calcium depletion was maximal (a condition achieved with an SR calcium channel opening drug) and partially when depletion was limited (a condition imposed by fatiguing stimulation, long-lasting depolarization, or low drug concentrations). With fluorescence and electron microscopic imaging we demonstrated massive movements of calsequestrin accompanied by drastic morphological SR changes in fully depleted cells. When cells were partially depleted no remodeling was found. The present results support the proposed role of calsequestrin in termination of calcium release by conformationally inducing closure of SR channels. A channel closing switch operated by calsequestrin depolymerization will limit depletion, thereby preventing full disassembly of the polymeric calsequestrin network and catastrophic structural changes in the SR. skeletal muscle | excitation/contraction coupling | cardiac muscle | muscle diseases | catecholaminergic polymorphic ventricular tachycardia T o start physiological contraction in striated muscles, a large amount of calcium moves from storage in the sarcoplasmic reticulum (SR) to the cytosol. Inside the SR, calcium is stored largely bound to calsequestrin, the only known protein dedicated to reversible ion buffering (1). In addition to storing calcium, calsequestrin has been proposed to mediate multiple functions (2). Although none of these proposed functions are proven, the likelihood that calsequestrin does more than buffering is supported by the widespread distribution of calsequestrin and calsequestrin-like orthologs in various cells, tissues, and organisms of Animalia and Plantae (reviewed in ref. 3), as well as the diversity of cardiac (4,5) and skeletal muscle (6,7) diseases linked to its mutations. Because the mutants show altered calcium release timing, calsequestrin is believed to modulate gating of ryanodine receptor (RyR) release channels. Because it forms long polymeric tendrils [also known as calcium wires (8)] ending near the SR channel mouth (9), calsequestrin is proposed to lead calcium toward the channel (8,10), a speculation that assumes (as first done in ref. 11) that reduction of dimensionality enhances diffusion.
Most importantly, calsequestrin is unique as a calcium buffer. Buffering power B, the incremental ratio of bound over free ligand concentration, is maximal at zero ligand concentration, when the buffer is ligand-free, for most buffers. The buffering power of calsequestrin measured in vitro instead increases with the concentrations of free and bound calcium (6). This increase in buffering ability occurs because the loci for most calcium binding are at the interfaces between calsequestrin monomers, requiring two monomers to achieve useful binding affinity (12). It follows that higher [Ca 2+ ] will promote polymerization, which, in turn, will provide additional sites for calcium binding. If adequately tuned, this cooperative feature will provide optimal calcium release ability at physiological [Ca 2+ ] SR (2).
[Ca 2+ ] SR decreases moderately (13) during normal skeletal muscle function; depletion is greater in cardiac muscle, where it reaches 50-90% (14). As crucial as polymerization appears to be for storage, and putatively other functions, its corollary, depolymerization, has never been demonstrated in a working cell upon depletion of the calcium store. Depolymerization would not just alter the buffer function of the protein; given the mechanical link of calsequestrin with other proteins in the couplon (15,16), including the calcium release channels of the SR (RyRs), depolymerization could be sensed as an allosteric signal by the RyRs, thereby providing a physical mechanism for their control. Depolymerization would derail the hypothetical diffusional enhancement by destroying the calcium wires; it might also allow migration of calsequestrin away from triads, with damaging targeting and functional repercussions. Finally, all calsequestrin mutations associated with diseases of skeletal muscle (6) and many of the known calsequestrin mutations linked to cardiac diseases (17) are known or predicted to alter its polymerization; therefore, it can be anticipated that any changes that hinder normal polymerization of this protein [including the posttranslational modifications recently studied (18)]

Significance
We show that calsequestrin, the main Ca 2+ storing protein of muscle, is polymerized inside the sarcoplasmic reticulum (SR) and its mobility increases greatly upon SR depletion, indicating depolymerization. Deep depletion causes massive calsequestrin migration and radical SR remodeling, often accompanied by a surge in intra-SR free Ca 2+ . The changes in calsequestrin polymerization observed in aqueous solutions therefore also occur in vivo. These changes help explain some uniquely advantageous properties of the SR as a source of calcium for contractile activation. The results support untested hypotheses about additional calsequestrin roles in the control of channel gating and facilitation of calcium flux. They also provide insights on the consequences of calsequestrin mutations for functional competence and structural stability of skeletal and cardiac muscle.
will have deleterious functional consequences. Whether and to what extent calsequestrin depolymerizes in situ, in working cells, is the central question of muscle function tested in the present work.
In single fibers of adult mouse muscle, we first characterize a kinetic feature of calcium release consistent with calsequestrin depolymerization and then evaluate the effects of calcium depletion on the intra-SR diffusional mobility of calsequestrin, which should increase upon depolymerization. We do it in myofibers of adult mice by imaging the fluorescence of tagged calsequestrin and its fluorescence recovery after photobleaching (FRAP) as a measure of the protein's mobility. Fluorescence tagging is achieved by fusion of calsequestrin (the skeletal isoform Casq1) with EYFP or with a calcium biosensor, which also provides a continuous readout of [Ca 2+ ] SR . Depleting interventions include drugs that open the SR calcium release channel or inhibit the sarco/endoplasmic reticulum calcium (SERCA) pump, and fatiguing electrical stimulation. EM imaging, probing possible structural changes in depleted muscles, completes the study.

Results
A Kinetic Hallmark of Calsequestrin Function. Fig. 1A (also Figs. S1 and S2) demonstrates a basic readout of calsequestrin function. The traces, reproduced from earlier work of our laboratory, are of calcium release flux elicited by a large depolarization imposed by patch-clamping; flux was derived from confocal images of [Ca 2+ ] cyto in wild-type (WT) mouse myofibers (blue) and a mouse null for both calsequestrin isoforms (red). As previously reported (19), the waveforms start with a sharp peak and then decay, rapidly in the calsequestrin-null model but going through a characteristic "hump" in the WT model. This feature strictly requires calsequestrin to be present (19) and can be restored by exogenous expression (Figs. S1 and S2). In all other experiments illustrated in Fig. 1A, the fast calcium buffer BAPTA [1,2bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid] was present in the internal solution to reduce calcium-dependent inactivation of the RyR channels (20). The presence of BAPTA increases the flux of calcium release and results in a sharper hump, which can be revealed using much briefer pulses, in turn, making elaborate kinetic studies feasible. Additionally, the sharply demarcated hump becomes quantifiable, as shown in Fig. 1C, in terms of the amount of released calcium on a time integral of calcium flux.
A kinetic feature of the hump illuminates its molecular mechanism. In Fig. 1 B-F, release flux is elicited by pairs of pulses. Flux in the second pulse is reduced, and its kinetics are modified. At very short intervals, thoroughly explored in other works (21,22), the peak of flux is absent. At intervals between 350 and 500 ms ( Fig. 1 B-E), the peak is largely recovered but the hump is absent. As shown in Fig. 1G, the peak recovers with a time constant τ of ∼200 ms, although it takes 10 s for the hump to show substantial recovery and more than 30 s for its full restitution. The red trace in Fig. 1G  These experiments (and others documented in Figs. S1 and S2) confirm that calcium released in the hump phase is supplied by calsequestrin. After this contribution, calsequestrin temporarily loses this function. Based on the requirement of calsequestrin polymerization for high-capacity calcium binding in vitro, we propose that this functional impairment reflects depolymerization of calsequestrin in situ, upon loss of calcium.

FRAP of Tagged Calsequestrin Reveals Changes in Its Diffusional
Mobility. If calsequestrin depolymerizes inside the depleted SR, it should become more mobile than in the rested muscle. This  prediction was tested by comparing FRAP of various luminal SR probes in myofiber bundles in resting and calcium-depleted conditions. Fig. 2A shows a small bundle of fibers from mouse flexor digitorum brevis (FDB) muscle inside the miniature chamber developed for this test. Slender bars attached eccentrically to rotary cylinders hold the bundle against the coverglass bottom, immobilized by stretch and N-benzyl-p-toluenesulfonamide (BTS). Fig. 2B shows a raw confocal image of the fluorescence of Casq1 fused with EYFP in cells of a calsequestrin-null mouse. EYFP-Casq1 fluorescence forms two bands per sarcomere, consistent with the location previously found for native Casq1 (23), and for Casq1 fused with D4cpV (24). The agreement indicates that Casq1 fused to either tag traffics in the same way as native Casq1, to SR terminal cisternae (TC). Each band of fluorescence therefore originates from the two TC in a triad, which cannot be resolved optically.
The dark rectangular areas shown in Fig. 2B were bleached by laser light at two different times: the one at the right while the fibers were at rest and the one at left 30 min later, 2 min after perfusion with depleting solution 2. A supporting time-lapse movie (Movie S1) illustrates the full experiment and demonstrates the performance of a custom program that automatically locates the bleached region for continued measurement in the presence of contractile or perfusion movements (of which the experiment is a most severe example). Fluorescence measured within the bleached areas is plotted in Fig.  2C. Recovery in the area to the right (region 1, open circles) was slow while the muscle was at rest and accelerated when the depleting solution was applied. Recovery in the area to the left (region 2, closed circles), bleached while in depleting solution, was visibly faster. Recovery is quantified in terms of the normalized fluorescence f ðtÞ ≡ FðtÞ=F ref . We show in SI Appendix and Fig. S3 that the initial rate of change of f ðtÞ, denoted as f ′ð0Þ and always given in units of (1,000 s) −1 , is proportional to the effective diffusion coefficient of the fluorescent species, D. We also establish the applicable value of the proportionality constant, which results in a D of 0.05 μm 2 ·s −1 when f ′ð0Þ equals 1. As shown in Fig. 2, f ′ð0Þ was 0.23 for the bleached region 1 and increased to 0.88 in region 2. The increase upholds the hypothesis that calsequestrin depolymerizes upon SR depletion (statistics are presented in Fig. 2F and Table 1).
Although f ′ð0Þ provides a measure of D immediately after bleaching, the "normalized rate of recovery" k e ðtÞ was used as a proportional measure of D beyond the initial times (Methods). With Fig. S4, we show that k e ðtÞ will decay monotonically with time if D is constant; instead, and as also shown in Fig. S4, it increased markedly in the first bleached region of Fig on Casq-null mice. Most, however, were done on WT mice; therefore, the exogenous Casq1 (which was expressed at 2-5 μmol/L cytosol) was less than 10% of the total [estimated at 50 μM or greater (25,26)]. The evolution of f ðtÞ in a representative case is shown in Fig. 2D ] SR was nearly zero). Meanwhile, the "immobile fraction" (determined as described in Methods) was reduced to zero, which implies that all of the tagged Casq1 became mobile. Bleaching in the four regions illustrated in Fig. 2D was achieved with the same amount of irradiation. Additional obvious evidence of the increase in mobility is the change in the value of f ð0Þ, the first recorded value of fluorescence after the bleaching irradiation is applied. The monotonic increase of f ð0Þ in the successively bleached regions (regions 1 to 4) reflects increased mobility, which resulted in greater recovery of fluorescence during the several seconds needed for irradiation and to switch from bleaching to the imaging mode. The same criterion is not applicable in cases (e.g., Fig. 2C) where the intensity of the bleaching irradiation was adjusted so that the comparison of rates was done at the same value of f ð0Þ.
The calcium biosensor D1ER, which is not fused to calsequestrin, was used as an alternative FRAP probe and [Ca 2+ ] SR reporter. In the example of Fig. 2E, f ′ð0Þ increased slightly upon partial depletion (region 2, black) and twofold when [Ca 2+ ] SR went under 100 μM (region 3, red). In the framework of the depolymerization hypothesis, the increase in mobility of D1ER indicates that the network of polymeric Casq1 hinders diffusion of proteins other than calsequestrin. The relative increase in diffusional mobility (D=D resting ) in a variety of conditions is summarized in Fig. 2F. Both central measures and distribution outliers of this variable had values greater than the null effect (D=D resting = 1). An alternative analysis, which compares f ′ð0Þ values pooled in different conditions, is shown in Table 1. From an average f ′ð0Þ of 1.37 (1,000 s) −1 , an estimate of 0.069 μm 2 ·s −1 is derived for D of D4cpV-Casq1 in cells at rest. This value is extremely low, even for molecules diffusing inside the endoplasmic reticulum/SR.
Other features of the change in mobility are illustrated in Fig.  3. Fig. 3A  Reversibility is a requirement for this change in Casq1, putatively depolymerization, to occur physiologically. We therefore checked whether the effects of depleting solutions on protein mobility were reversible. Fig. 3B graphs FRAP in a fiber expressing EYFP-Casq1 (open circles). The FRAP rate f ′ð0Þ (i.e., the slope of the line determined by the open symbols) was initially low, as usual in a fiber at rest; it increased greatly upon exposure to solution 2 and appeared to decrease rapidly on washout. The evolution is best shown with the "normalized rate" k e ðtÞ (blue squares).
The return to low protein mobility required recovery from calcium depletion. This recovery is shown in the experiment of Fig. 3C, which consisted of measuring FRAP rate of D4cpV-Casq1 in six regions, bleached successively in the same muscle bundle under different conditions. The bundle was exposed to the reversible SERCA inhibitor 2.5-di(tert-butyl)-1,4-hydroquinone (TBQ) during the time marked by the blue bar, causing [Ca 2+ ] SR to decrease slowly, whereas f ′ð0Þ increased, up to fourfold after 2 h. The changes reversed partially upon washout.
Two other observations were indicative of calsequestrin depolymerization. Depleting solutions often produced an oscillation in [Ca 2+ ] SR (examples are shown in Fig. 3D and Fig. S5). As shown in Fig. 3D, the oscillations started with a decay, followed by an upswing that surpassed the resting [Ca 2+ ] SR value in some cases and a final fall to near zero. As illustrated in Fig. 3D and Fig. S5, the magnitudes of the upswing and initial decay were correlated, and both increased with the dose of 4-CMC. A large increase in [Ca 2+ ] SR while calcium release channels are open can only be due to a large decay in SR calcium buffering power, such as the decay that follows full depolymerization of Casq1. The association between the magnitude of the upswing in [Ca 2+ ] SR and its preceding dip suggests that the depletion is the cause of the polymeric network's collapse.
In nearly every experiment, a fraction of the bleached fluorescence was unrecoverable (examples are shown in Figs. 2 C and D and 3 B and D). As documented in the collective plot of Fig. 3E, this fraction, which presumably represents Casq1 fully immobilized in the polymeric lattice, was large at high [Ca 2+ ] SR and decreased with depletion, which is evidence of a weaker retention of Casq1 in the lattice. ] SR as determined by the biosensor at the time of bleaching. Column 5 gives numbers of cells used in the measurements. n.a., not available. *Significance of the effect of the depleting intervention at P < 0.05 in a two-tailed t test of differences of averages. Fig. 2F provides an alternative comparison of effects, as differences paired in individual fibers.
† No biosensor was present.
These observations indicate that up to 60% of the tagged protein is immobile in the time scale of our experiments in cells at rest, and that part of it becomes mobile as [Ca 2+ ] SR is reduced. When the depletion is slow in onset, the increase in mobility is gradual and reversible. When the depletion is fast, the oscillation in free [Ca 2+ ] SR suggests widespread collapse of the calsequestrin network, leading to sudden loss of calcium buffering capacity and a corresponding surge in [Ca 2+ ] SR .
Electrical Stimulation Increased Calsequestrin Mobility. SR depletion was also induced by "fatiguing" field stimulation patterns, confirmed to cause mechanical fatigue in separate experiments (Fig.  S6). Fatiguing stimulation achieved a reduction in [Ca 2+ ] SR of 30-40% in 20 min of application and caused a twofold to threefold increase in f ′ð0Þ. However, for the reduction in [Ca 2+ ] SR to be sustained while FRAP was measured, stimulation had to continue, often resulting in movement. In alternative experiments, a low dose of the SERCA inhibitor cyclopiazonic acid (CPA; 0.5 mM) was added to Tyrode. Fatiguing stimulation then caused a decrease in [Ca 2+ ] SR that was larger (nearly 70% of resting value) and sustained, so that stimulation could be stopped during FRAP measurement. The increase in calsequestrin mobility in these tests was similar to the increase in mobility induced with depleting drugs (D=D resting = 3.3; Fig. 2F).
Displacement of Calsequestrin Demonstrated by Imaging. Fig. 4 shows D4cpV-Casq1 fluorescence in the muscle of a WT mouse before and after chemical depletion, together with EM images of the same muscle. In Fig. 4A is one member of a "z-stack," consisting of 30 images at 100 nm of vertical separation. The numerical aperture of the objective, voxel dimensions, and confocal pinhole radius were all set to optimize spatial resolution. Fig. 4B is the result of deblurring Fig. 4A (a procedure that uses all members of the stack), and Fig. 4C is a 3D rendering of the deblurred stack. The images confirm the observation (24) that in cells at rest, D4cpvtagged Casq1 resides mostly in two bands per sarcomere. Fluorescence costaining (23, 24) has shown that the bands correspond to triads, each with two TC not resolved in these images. The narrow and wide spaces between these calsequestrin bands correspond respectively to the I and A bands of the sarcomere (24).
The effect of deep depletion is demonstrated in Fig. 4D, a 3D rendering of a z-stack acquired on the same fiber after exposure to solution 5. The treatment, which reduced [Ca 2+ ] SR to between 0 and 20 μM in three experiments, caused a major relocation of the fluorescent species into largely longitudinal structures.
The nature of these structures was explored in EM images of sections of the same depleted muscles. Fig. 4E, from a muscle at rest expressing the biosensor, shows the normal arrangement of triads (indicated here and elsewhere by red arrows). Calsequestrin is visible as the granular content of the two TC in a triad (27,28) and is mostly absent in longitudinal SR (indicated by the yellow arrow in Fig. 4E). In cross-sections (Fig. 4F), the longitudinal SR is largely empty (yellow arrow), but it occasionally exhibits calsequestrin-like granular content as well (green arrow), particularly at the level of the I band as illustrated here. Fig. 4 G and H show longitudinal and transversal sections from the muscle in Fig.  4 A-D, fixed after depletion. The images in Fig. 4 G and H show intermyofibrillar spaces occupied by a labyrinthine array of membranes with an occasional granular content, presumably representing polymerized Casq1 (yellow arrows; details are shown in Fig. 4 I and J). The array of membranes includes denser areas, often in pairs, which are recognizable by size and placement as remnants of triads, devoid of their transverse tubule (red arrows in Fig. 4 G and J). The presence of these elements identifies the intermyofibrillar array of membranes as displaced, morphologically altered SR. By shape, size, and orientation, they correspond to the longitudinal structures in the fluorescence images. Because it is visible by fluorescence, the content in these structures must include D4cpV-Casq1. As explained in an earlier section, the exogenous Casq1 is, on average, not more than 10% of the endogenous Casq1; therefore, the massive displacement of volume and content from triadic toward intermyofibrillar structures must involve the native Casq1 as well. The intricacy of the SR structures in the depleted cells gives the impression of growth in SR area. To test this possibility, we computed the ratio of intersected SR membrane length and image area, which is equivalent to SR membrane area density [area per unit volume (29)]. Details of the comparison are provided in Fig.  S8. The SR area density was smaller in the depleted condition, albeit not significantly so. The result of the comparison suggests redistribution of SR membrane upon deep Ca 2+ depletion.
The structural consequences of milder depleting solutions are documented in Fig. 5. Fig. 5 A-D compares fluorescence in the same fiber before (Fig. 5 A, a-D, a) and after (Fig. 5 A, b-D, b) treatment with solution 2, which caused a reduction in [Ca 2+ ] SR to less than 40% of its resting value. Fig. 5 A and B shows unprocessed z-stacks. Fig. 5 A, a-B, b shows collective views of the stacks, generated by overlapping all sections. Fig. 5 A, b and B, b shows that the treatment results in blurring of the distribution of fluorescence, which moves slightly away from triads. Fig. 5 C and D, comparing the effects of exposures of different duration to the same solution, shows that the displacement becomes more evident as exposure time increases.
The effects of this milder treatment were also explored with electron microscopy. Fig. 5E illustrates a WT myofiber at rest, expressing D4cpV-Casq1. Two TC in the image (Fig. 5E, red arrow) have granular content, previously identified as polymeric calsequestrin, whereas longitudinal SR elements appear empty or have unstructured content (Fig. 5E, yellow arrow). Fig. 5 F-H is from different muscles fixed after 25 min in solution 2; these images illustrate the observation of Casq1-like granular content in longitudinal SR (arrows), which, however, was not very frequent. The occasional presence of such content in extratriadic SR of untreated muscle, plus the abundance of SR elements with content of ambiguous density and texture (Fig. S7), made it impossible to decide, based on EM images alone, whether the depleting treatment caused a significant increase in extratriadic Casq1.
The structural effects of fatiguing stimulation are illustrated in Fig. 6. As stated before, stabilizing the depletion due to tetanic activity required stimulating in the presence of CPA. Images in Fig. 6 compare the same myofiber before (Left) and after (Right) stimulation. The pattern changed subtly, from having two triadic bands per sarcomere at rest (as in Fig. 5 A, a-D, a) to a more diffuse appearance consistent with displacement of the protein. In these and other images obtained similarly, the displaced calsequestrin appears to occupy preferentially the narrow space between triadic bands, marked by a yellow arrow in Fig. 6 A, a and  b, a space known to correspond to the sarcomeric I band (24).
EM images of myofibers fixed after fatiguing stimulation (shown in Fig. S8)   distinction, we compared systematically images of transversal sections (as in Fig. 4 F and H), but we did not find significant differences in the frequency of "filled" SR elements between resting and stimulated muscles.

Discussion
Being a calcium-binding protein, calsequestrin might have a number of signaling roles; its sole demonstrated function in cells, however, is reversible calcium storage (19,25). Calsequestrin's unique buffer properties appear well-suited for this role; they include a high calcium-binding capacity and a buffering power B that increases at increasing [Ca 2+ ], as reflected in a binding curve of positive curvature (or cooperativity > 1). In studies in vitro, the increase in B occurs in the same range of [Ca 2+ ] that induces polymerization (6,30), which implies that polymerization is crucial for effective buffering. The potential implications for physiology of this property (reviewed in ref. 2) have been the subject of much speculation, all of which remains untested.

Calsequestrin Undergoes Major Conformational Changes upon Calcium
Release. Given that Casq1 provides much of the calcium that signals muscle contraction (25,26), the buffering features found in solution should transfer to the storage properties of the SR in cells. Indeed, strongly cooperative calcium buffering by the SR was first documented by Pape et al. (31), who showed with frog muscle that the calcium buffering power of the empty SR is negligible compared with the value at rest, a property they attributed to calsequestrin. Working on mouse muscle, Royer et al. (19) confirmed the attribution, quantified the contribution by Casq1 at 75% of released calcium (26), and equated it to the hump of calcium flux caused by a large depolarization (19). Here, we characterized this component further (Fig. 1), showing that it becomes unavailable at short times after one occurrence and that its restoration proceeds more slowly than recovery of free [Ca 2+ ] SR , and is therefore not limited by it. The delay implies that Casq1 undergoes major changes in conformation when cycling calcium on and off, and that the off change disables its function temporarily. The observation is consistent with the notion that calsequestrin must polymerize to achieve high-capacity calcium binding and depolymerize as it loses bound calcium. ] SR , also evinces a causal link between depletion and the increase in mobility.

The Changes in Calsequestrin Mobility Are Explained by Its
Depolymerization. Both the changes in mobility and the exhaustion and delayed recovery of the hump of calcium release can be justified simply as consequences of depolymerization of Casq1 upon [Ca 2+ ] SR depletion. To begin with, Casq1 appears to be polymerized in the cell at rest. Mean initial rates of recovery, f ′ð0Þ, of D4cpV-Casq1 fluorescence were 0.63 and 1.37, respectively, in WT and calsequestrin-null mice ( Fig. 3F and Table 1). A value of 1 for f ′ð0Þ corresponds to D = 0.05 μm 2 ·s −1 (SI Appendix and Figs. S3 and S4). This value is extremely low in comparison to other  large molecules inside the endoplasmic reticulum [e.g., 0.5 μm 2 ·s −1 for elastase-GFP (32), reviewed in ref. 33]. The low estimates of D and the observation that tagged Casq1 is partly immobile agree with the consensus that Casq1 constitutes the dense polymeric network visible inside TC in EM images (9,27). The detailed similarity in structure of calsequestrin polymers in crystals (12) and in vivo (34) provides additional evidence and supports the extrapolation to cells of calsequestrin properties observed in solution, including the close association between polymerization and calcium binding.
That the increase in probe mobility upon [Ca 2+ ] SR reduction is accompanied by a decrease of the immobile fraction of fluorescence (Fig. 3E), that it takes longer than the changes in [Ca 2+ ] SR (Fig. 3C), and that it appears to accelerate at [Ca 2+ ] SR below a threshold (Fig. 3A) are all indications that the polymeric network weakens and eventually comes apart upon depletion. That the mobility of another protein, D1ER, is also increased (Fig. 2 E  and F) shows that a barrier to diffusion of all large molecules is lowered when [Ca 2+ ] SR falls, consistent with weakening of a structure that fills the TC.
A large-scale change in the calsequestrin lattice upon major depletion also explains the production of [Ca 2+ ] SR upswings (presumably by calcium that massively leaves the calsequestrin network as it disassembles) and specifically justifies the observed correlation between amplitude of intra-SR transients and strength of the depleting stimulation. Intra-SR calcium transients were first reported in vesicular SR fractions (35) and studied in detail in the rat (36). With simultaneous measurements of FRAP rate and [Ca 2+ ] SR , we now show that they coexist with an increase in Casq1 mobility, confirming that they are associated with and, in all likelihood, due to changes in the polymeric network of Casq1, as proposed by Launikonis et al. (37).
Tagged Casq1 Moves Away from TC upon Depletion. The change in the mobility and polymerization state of Casq1 should allow its movement away from TC into extratriadic SR. Such movement was observed in images of muscles depleted pharmacologically (Figs. 4 and 5). The EM images of SR structures, in which polymerized Casq1 is visible as dense granular content, demonstrated massive movement in muscles treated with high 4-CMC (e.g., Fig. 4G). In these muscles, the gross morphology of the SR changed, in agreement with the notion that polymerized calsequestrin is a determinant of normal SR morphology in skeletal and cardiac muscle (38,39).
There were no obvious changes in EM images of muscles subjected to the milder depleting solution 2, even though the fluorescently tagged Casq1 was displaced (Fig. 5). The discrepancy may reflect a requirement of polymerization for calsequestrin to be visible in EM images. EM visualization of displaced calsequestrin may therefore be limited to conditions in which dissociated calsequestrin migrates away from triads and reconstitutes polymers elsewhere or the polymeric network breaks into pieces that can diffuse, conditions only met in muscles treated with the higher dose of 4-CMC. Such rearrangements are probably irreversible and unlikely to occur physiologically.
Fatiguing Stimulation Causes Minor Displacement of Casq1. Tetanic stimulation caused moderate, rapidly recovering SR depletion. Mobility increased in these cases, but its evaluation required maintaining the stimuli during FRAP, which elicited contractions and required continued perfusion, both sources of artifact. A greater increase in mobility was recorded after electrical stimulation in the presence of CPA, which made the depletion sustained, so that FRAP rates could be determined without movement artifact. The magnitude of the increase in mobility was then comparable to the magnitude caused by chemical depletion (Fig. 2F). By contrast, a systematic examination of EM images of transversal sections of electrically stimulated muscles found no changes in the dense content of longitudinal SR elements, identified as polymeric calsequestrin. This negative result is consistent with our failure to demonstrate displacement of dense content in EM images of cells subjected to mild depleting solutions and may again reflect the impossibility to visualize monomeric calsequestrin by EM. Therefore, we conclude that the Casq1 network is not grossly disassembled by fatiguing stimulation as applied here.
Findings in vitro help interpret the present observations; indeed, it has been shown that the crystal lattices of calsequestrin and their asymmetrical units establish order of longer range as [Ca 2+ ] increases (18). Thus, upon moderate alteration of [Ca 2+ ] and [K + ], order may decrease or increase in a gradual manner, without total breakdown or reconstruction of the basic lattice. Extrapolating to cells, and in line with the observed limited displacement of Casq1 and the increase in its mobility, we propose that limited changes in [Ca 2+ ] SR weaken the Casq1 network, without full disassembly.
The argument for gradual change, rather than a binary alternative of polymerization or depolymerization, can also be cast in terms of diffusion with binding. If the binding affinity K of Casq1 to the fixed polymeric network decreases in a continuous manner as [Ca 2+ ] is reduced, the effective diffusion coefficient of Casq1, D eff = D=ð1 + KÞ, will increase gradually as [Ca 2+ ] and K decrease. Depletion of calcium in the SR is believed to contribute to muscle fatigue because it reduces the [Ca 2+ ] gradient driving calcium release, possibly reducing contractile activation (e.g., ref. 40). The present results suggest a second mechanism: The changes in the Casq1 network induced by the mild depletion associated with fatiguing activity will reduce the buffering power of the SR, and thus contribute to the reduction in calcium flux.
A Self-Protecting Calsequestrin Switch on Calcium Release? Sztretye et al. (20) showed that long-lasting membrane depolarization in myofibers under voltage clamp causes only partial calcium store depletion; their finding is in agreement with the present evidence that depolymerization may be graded and remains limited during normal muscle function. The same work showed that depletion is partial because RyR channels close before the SR is fully depleted. The closure may be mediated by Casq, because in Casqnull muscles, RyRs stay open, allowing for full depletion. Here, we show that complete depletion often leads to abrupt, large-scale depolymerization (as reflected in the oscillations in [Ca 2+ ] SR observed under these conditions; Fig. 3D and Fig. S5) and, ultimately, to major alterations in SR structure (Fig. 4). If, as proposed by Sztretye et al. (20), changes in the Casq1 lattice due to partial SR depletion constitute the conformational signal that close RyR channels, the ultimate benefit of this modulation would be to prevent both full emptying of the calcium store and catastrophic disassembly of the Casq1 network.
Polymerization Defects of Calsequestrin as Causes of Disease. Studied in vitro, two disease-linked variants of Casq1, D244G (7) and M87T (41), show altered polymerization, which requires higher [Ca 2+ ], leads to abnormal polymeric structure, and reduces calcium binding in solution in both cases (6). The present evidence that Casq1 undergoes depolymerization in vivo supports the view that muscle weakness in patients with the D244G mutation is due to reduced calcium storage and release (6,42). Additionally, changes in the structure of the polymer (6) may explain the abnormal assembly of SR components characteristic of "surplus protein" or "vacuolar aggregate" myopathies (43,44).
Although M87T is not causative of disease per se, it increases penetrance of RyR1 mutations linked to malignant hyperthermia (MH) (41), a disease characterized by events of uncontrolled calcium release that cause hyperpyrexia and rhabdomyolysis. We now show that conditions favoring RyR channel opening and rapid calcium release cause oscillations in [Ca 2+ ] SR , presumably due to disassembly of the Casq1 lattice. As reported by Cully et al. (36), these [Ca 2+ ] SR transients are regenerative; therefore, they may be a factor, as yet unrecognized, precipitating the MH event. Variants that destabilize the Casq1 lattice, including M87T (41), could aggravate the MH condition by this means.
The present findings have implications for cardiac physiology and disease. Depolymerization should be more complete in the heart, because the SR appears to deplete more fully there than in skeletal muscle (e.g., refs. 45,46). Because TC are smaller in cardiac cells and connect more sparsely to longitudinal SR (47), the loss of calsequestrin (Casq2) from these TC would be limited and easier to reverse, even after full disassembly of the polymeric network. The present findings therefore support the idea that Casq2 detachment from cardiac couplons constitutes a signal for closing RyR2 channels and termination of SR calcium release (reviewed in ref. 48).
Disrupted polymerization has been demonstrated for three missense mutations of Casq2. Mutations R33Q and D307H, associated with a severe form of catecholaminergic polymorphic ventricular tachycardia, yield monomers that form an altered dimer, whereas mutant L167H has a disrupted hydrophobic core in an internal domain, resulting in random aggregates that do not respond to varying calcium concentration (17). These alterations will interfere with beat-to-beat cycling of calcium. Additionally, by interfering with Casq2 trafficking and retention in TC (49), weaker polymerization may explain the reduction in Casq2 content that accompanies most of its mutations (4). If, as suggested by the present results, polymerization and depolymerization are dynamic processes that repeat with every heartbeat, any hindrance to these processes will have repercussions on calsequestrin endowment and structural stability of the cardiac SR.
In sum, the present work establishes calcium-dependent polymerization and depolymerization of calsequestrin as a critical hub for modulation of function, trafficking, correct couplon assembly, stability of SR structure, and their alterations in human disease (50)(51)(52).
Animal use protocols were found consistent with ethical standards and approved by the Institutional Animal Care Use Committee of Rush University. Transfection and isolation of muscle followed methods adapted from DiFranco et al. (57), and described by Royer et al. (19). Four to seven days after transfection, the animals were killed by CO 2 inhalation, and FDB muscles were removed for imaging or functional studies. In experiments with single cells, the methods of cell separation and patch voltage clamping were as described (58).
Calcium Concentration Measurements. Cytosolic transients were recorded in isolated cells under whole-cell patch-clamp with rhod-2 introduced via the patch pipette. The [Ca 2+ ] cyto and calcium release flux were derived by conventional procedures (58) from line scan and x-y confocal images (SP2 acousto-optical beam splitter; Leica). Single-cell flux measurements were performed at 20-22°C in "external" solution. The [Ca 2+ ] SR was determined from suitable Förster resonant energy transfer (FRET) ratios, from which [Ca 2+ ] SR was derived as published (24).
FRAP Measurements. FRAP measurements were done in thin bundles dissected from FDB muscles and mounted in a custom 3D-printed chamber ( Fig. 2A; patent PCT/US16/26097) designed to fit inside a 35-mm thermostatic jacket. Miniaturization, flow streamlining, and a built-in inlet manifold, made possible by the 3D printing technique, accomplish rapid, turbulence-free perfusion and solution exchange. Thin stainless wires attached eccentrically to rotary cylinders allow stretching and pressing the muscle bundle to the coverglass chamber bottom for immobilization (completed by BTS in the perfusing solution) and imaging with high numerical aperture objectives. Trains of stimuli applied with built-in platinum electrodes under continuous perfusion produce a measured current of steady amplitude. The chamber can be built to order (Department of Molecular Physiology and Biophysics, Loyola University, Chicago; available upon request from pjcaron@lumc.edu).