Junctional Cleft [Ca²⁺]_i Measurements using Novel Cleft-Targeted Ca²⁺ Sensors

Abstract

Rationale

Intracellular Ca²⁺ concentration ([Ca²⁺]_i) is regulated and signals differently in various subcellular microdomains, which greatly enhances its second messenger versatility. In the heart, sarcoplasmic reticulum (SR) Ca²⁺ release and signaling is controlled by local [Ca²⁺]_i in the junctional cleft ([Ca²⁺]_Cleft), the small space between sarcolemma and junctional SR. However, methods to directly measure [Ca²⁺]_Cleft are needed.

Objective

To construct novel sensors that allow direct measurement of [Ca²⁺]_Cleft.

Methods and Results

We constructed cleft-targeted [Ca²⁺] sensors by fusing Ca²⁺-sensor GCaMP2.2 and a new lower Ca²⁺-affinity variant GCaMP2.2Low to FKBP12.6, which binds with high affinity and selectivity to ryanodine receptors (RyRs). The fluorescence pattern, affinity for RyRs and competition by un-tagged FKBP12.6 demonstrated that FKBP12.6-tagged sensors are positioned to measure local [Ca²⁺]_Cleft in adult rat myocytes. Using GCaMP2.2Low-FKBP12.6, we showed that [Ca²⁺]_Cleft reaches higher levels with faster kinetics than global [Ca²⁺]_i during excitation-contraction coupling. Diastolic SR Ca²⁺ leak or sarcolemmal Ca²⁺ entry may raise local [Ca²⁺]_Cleft above bulk cytosolic [Ca²⁺]_i ([Ca²⁺]_Bulk), an effect that may contribute to triggered arrhythmias and even transcriptional regulation. We measured this diastolic standing [Ca²⁺]_Cleft–[Ca²⁺]_Bulk gradient using GCaMP2.2-FKBP12.6 vs. GCaMP2.2, using [Ca²⁺] measured without gradients as a reference point. This diastolic difference ([Ca²⁺]_Cleft=194 nmol/L vs. [Ca²⁺]_Bulk=100 nmol/L) is dictated mainly by the SR Ca²⁺ leak, rather than sarcolemmal Ca²⁺ flux.

Conclusions

We have developed junctional cleft targeted sensors to measure [Ca²⁺]_Cleft vs. [Ca²⁺]_Bulk, and demonstrated dynamic differences during electrical excitation and a standing diastolic [Ca²⁺]_i gradient which could influence local Ca²⁺-dependent signaling within the junctional cleft.

INTRODUCTION

Ca²⁺ is a universal second messenger involved in activation/regulation of cellular processes as diverse as neurotransmitter release, muscle contraction, metabolism, hypertrophic signaling, and cell death. To fulfill such diverse roles and yet be specific enough to trigger distinct responses, information is encoded in the amplitude, temporal properties, and subcellular localization of Ca²⁺ signals. It is now largely recognized that Ca²⁺ is regulated and signals differently in various subcellular microdomains.^1-3 For example, GABA inhibition in hippocampal neurons results in spatially confined inhibition of Ca²⁺ transients shortly after a back-propagating action potential, which may play a key role in regulation of synaptic plasticity.⁴

In the heart, Ca²⁺ release from the sarcoplasmic reticulum (SR) is a key event for excitation-contraction coupling (ECC), ventricular arrhythmias, mitochondrial function and nuclear transcription.⁵ Cardiac SR Ca²⁺ release is controlled by the local [Ca²⁺] in the junctional cleft ([Ca²⁺]_Cleft), the small, restricted space between the apposing sarcolemma (mostly T-tubules) and junctional SR membrane. Voltage-gated L-type Ca²⁺ channels (LTCC) and Ca²⁺-activated Ca²⁺ channels (or ryanodine receptors, RyRs) are located at these junctions in the sarcolemma and SR membrane, respectively, and are essential in this local control.^6-8 During electrical excitation, [Ca²⁺]_Cleft is expected to rise higher and faster than in the bulk cytosol.^9-12 Through Ca²⁺-dependent inactivation, [Ca²⁺]_Cleft feeds back on LTCC and thus plays an important role in shaping the action potential waveform. Several studies^12-17 suggest that the Na⁺/Ca²⁺ exchanger (NCX), the main route for Ca²⁺ extrusion in cardiac myocytes, also has access to the cleft and influences [Ca²⁺]_Cleft during both diastole and ECC. In diastole, NCX may help to keep [Ca²⁺]_Cleft low and limit spontaneous RyR activation. Indeed, during stochastic RyR opening events, Ca²⁺ removal by NCX may limit local Ca²⁺ release within a given RyRs cluster, which limits Ca²⁺ sparks and waves, favoring smaller, non-spark Ca²⁺ release events.¹⁸ During an action potential and rapid Na⁺ influx, local NCX may boost [Ca²⁺]_Cleft and the efficacy of local SR Ca²⁺ release, especially in the latent period before LTCC opening.¹⁹

[Ca²⁺]_Cleft could exceed that in bulk cytosol ([Ca²⁺]_Bulk) even during diastole. This is because spontaneous SR Ca²⁺ release (leak) and Ca²⁺ influx caused by stochastic openings of LTCCs may create a standing [Ca²⁺] gradient between the cleft (where Ca²⁺ influx and release happen) and the bulk cytosol (where most of SR Ca²⁺ uptake occurs). Indeed, there is functional evidence for elevated [Ca²⁺]_Cleft in NCX knockout cardiac myocytes, that causes reduced LTCC availability and higher ability of the remaining LTCC to trigger SR Ca²⁺ release.²⁰

Despite its importance, methods to directly measure [Ca²⁺]_Cleft are only beginning to be tested.^21,22 Indirect estimates based on the dynamic properties of Ca²⁺-dependent inactivation of LTCC^12,23 and NCX current^12,24 during electrically-triggered SR Ca²⁺ release suggest that [Ca²⁺]_Cleft peaks within 10-20 ms and reaches tens or even hundreds of μM. In contrast, the global Ca²⁺ transient reaches a maximum of approximately 1 μM within ~70-80 ms.⁶ While strongly supporting the theory of local control of ECC, such measurements are indirect, difficult to calibrate, depend on the specific membrane localization of these transporters, and provide no information on diastolic [Ca²⁺]_Cleft. Our aim was to construct novel Ca²⁺ sensors that are specifically targeted to the sarcolemma-SR junctions and thus report local [Ca²⁺]_Cleft. We used these sensors to assess the dynamic changes in [Ca²⁺]_Cleft during ECC and to demonstrate the existence of a standing [Ca²⁺] gradient during diastole between the cleft and bulk cytosol.

METHODS

Detailed methods are included in the Online Supplement

Mutagenesis of GCaMP2-based Ca²⁺ sensors, cleft-targeting, plasmid construction, and expression in adenoviral vectors

The genetically encoded Ca²⁺ sensors (GECI) GCaMP2.2 and GCaMP2.2Low were constructed by introducing the T203V and, respectively both T203V and D133E mutations in GCaMP2. For targeting to the junctional cleft, GCaMP2.2 and GCaMP2.2Low were fused to the N-terminus of FKBP12.6 through a peptide linker. The GCaMP2.2 constructs were subcloned into the pshuttleCMV vector using Bgl II and Not I restriction sites. Subsequent adenovirus generations and amplifications were done using the Adeasy system (Agilent).

Protein expression and purification

Tagged and untagged GCaMP2.2(Low) sensors were cloned into a pRSET expression vector and amplified in BL21 Star (DE3) pLysS cells, then purified using Profinity IMAC Ni Charged Resins (Bio-Rad) and further subjected to size exclusion chromatography.

Ventricular myocyte isolation, culture and adenoviral infection

All animal protocols were approved by the animal welfare committee at UC Davis and conform to the NIH Publication No. 85-23 (revised in 1996). Rat ventricular myocytes were isolated by digestion with collagenase, cultured in supplemented M199 media and infected with adenoviruses expressing one of the GCaMP2.2 sensors. For some experiments, freshly isolated myocytes were permeabilized with 50 μg/mL saponin for 3 minutes. All experiments were done at room temperature.

Fluorescence measurements

GCaMP2.2s fluorescence was recorded with a Live-5 laser scanning confocal microscope (excitation=488 nm, emission>505 nm). Ca²⁺ transients were measured in linescan mode (2 ms/line). In other experiments, 2D images were taken 1 sec apart.

Statistical analysis

The statistical differences between groups were determined using the student’s t-test.

RESULTS

Low affinity GCaMP-based genetically-encoded Ca²⁺ indicator

GCaMPs are GECIs consisting of a circularly permutated, enhanced GFP that is flanked by calmodulin (CaM) at the C-terminus and by the CaM binding peptide myosin light-chain kinase M13 at the N-terminus (Fig. 1A), and have been used to detect [Ca²⁺] in vivo.^25-28 The T203V mutation increases the brightness and dynamic range of GCaMP2 (GCaMP2.2),²⁹ and we used this as a starting point to screen mutations that decrease the Ca²⁺ binding affinity of GCaMP2.2, while maintaining brightness and dynamic range. EF-hand loop mutational screening identified GCaMP2.2Low, which contains a D133E mutation³⁰ in the EF-hand loop IV of calmodulin (Fig. 1A). GCaMP2.2Low has a similar baseline fluorescence and dynamic range as GCaMP2.2, but its affinity for Ca²⁺ binding is ~10-fold lower than that of GCaMP2.2 (K_d≈5 μmol/L vs. 450 nmol/L, Fig 1B).

Figure 1. Development and validation of Ca²⁺ sensors that report local [Ca²⁺]_Cleft.

(A) Schematic representation of GCaMP2.2 and GCaMP2.2-FKBP12.6 sensors. The mutation (D133E) indicated in the Top Panel was introduced to obtain GCaMP2.2Low. (B) Ca²⁺-dependence of purified GCaMP2.2, GCaMP2.2-FKBP12.6, GCaMP2.2Low and GCaMP2.2Low-FKBP12.6 in saline solution. (C) Confocal images of saponin-permeabilized rat ventricular myocytes that were incubated for 10 min with 100 nmol/L purified GCaMP2.2 (a) or GCaMP2.2-FKBP12.6 (b). In c, permeabilized myocytes were first incubated with 10 μmol/L non-fluorescent FKBP12.6 for 5 min, followed by exposure to 100 nmol/L GCaMP2.2-FKBP12.6 in the continued presence of FKBK12.6. The full cell images related to a-c are shown in Online Fig. I. (D) The time course of GCaMP2.2-FKBP12.6 wash-in and wash-out in permeabilized myocytes. We monitored the increase in cell fluorescence upon adding 100 nmol/L GCaMP2.2-FKBP12.6 to the external solution and the decrease in fluorescence following the withdrawal of GCaMP2.2-FKBP12.6. Data are mean ± SEM of 8 separate experiments. These measurements were used to calculate the rate constants of GCaMP2.2-FKBP12.6 binding (k_on) and unbinding (k_off) to/from RyRs shown in Fig. S2. (E) Ca²⁺-dependence of the four sensors measured in permeabilized rat ventricular myocytes. To prevent contracture of the cells at elevated [Ca²⁺], all solutions contained 20 mmol/L BDM. All experiments were done at room temperature.

GCaMP2.2-FKBP12.6 Ca²⁺ sensors report [Ca²⁺]_Cleft in the space between LTCC and RyR

Our strategy for targeting Ca²⁺ sensors to a specific subcellular microdomain (the cleft) is to attach GCaMP2.2 and our new GCaMP2.2Low to a molecule that specifically targets to that microdomain. Based on their Ca²⁺ affinities (Fig 1B), GCaMP2.2 is useful for measuring diastolic [Ca²⁺]_i while GCaMP2.2Low could record the large Ca²⁺ transients expected in the junctional cleft. To target GCaMP2.2s to the junctions, we fused them to the N-terminus of FKBP12.6 (Fig. 1A). FKBP12.6 is an endogenous protein that binds with high affinity (K_d~1 nmol/L), specificity and 1:1 stoichiometry to RyR2 monomers, but does not greatly influence RyR2 function.³¹ Our prior work showed that fluorescent-tagged FKBP12.6 binds in a striated pattern along T-tubules at the sites of Ca²⁺ spark initiation, and that binding is prevented by preincubation with non-fluorescent FKBP12.6.³¹ Linking FKBP12.6 to the GCaMP sensors does not significantly affect their Ca²⁺ affinity (Fig. 1B) or Ca²⁺ binding kinetics (Online Table I). Saponin-permeabilized myocytes exposed to 100 nmol/L purified untagged GCaMP2.2 show uniform cytosolic distribution of the sensor (Fig 1Ca; Online Fig. I shows the entire cell). In contrast, GCaMP2.2-FKBP12.6 has a striated fluorescence pattern (Fig 1Cb) like that seen for purified FKBP12.6 tagged with a small molecule fluorophore,³¹ which suggests that the sensor is appropriately localized at the z-line where RyRs and clefts are situated. To further test the specificity of GCaMP2.2-FKBP12.6 binding at RyRs, we exposed permeabilized myocytes to GCaMP2.2-FKBP12.6 in the presence of excess non-fluorescent FKBP12.6 to saturate RyRs (Fig 1Cc). In this case, GCaMP2.2-FKBP12.6 shows uniform cytosolic fluorescence. These data suggest that the FKBP12.6-tagged GCaMP2.2 sensors bind with high selectivity to RyRs, which are mainly cleft-localized.

We next assessed the affinity of GCaMP2.2-FKBP12.6 binding to RyRs by performing sensor wash-in and wash-out experiments in permeabilized myocytes (Fig 1D). The rate constants for the dissociation (k_off) and association (k_on) of the sensor from/to RyRs were calculated from exponential fits of the decline in GCaMP2.2-FKBP12.6 fluorescence upon removing the sensor from the solution (k_decline=k_off) and the GCaMP2.2-FKBP12.6 fluorescence increase upon addition of the sensor (k_increase=k_on•[GCaMP2.2-FKBP12.6]+k_off) (Online Fig. II). The dissociation constant of the RyR - GCaMP2.2-FKBP12.6 complex (K_d) was calculated as k_off/k_on and yielded a value of 15.5±3.3 nmol/L. A similar calculation showed that GCaMP2.2Low-FKBP12.6 binds to RyRs with a K_d of 43.8±6.3 nmol/L. While these affinities are lower than previously reported for FKBP12.6 alone (~1 nmol/L),³¹ GCaMP2.2-FKBP12.6 and GCaMP2.2Low-FKBP12.6 bind RyR with relatively high affinity.

We then calibrated the [Ca²⁺]-dependence of the sensors in the myocyte environment (Fig. 1E). The dynamic range of the sensors in myocytes was comparable to that measured in vitro (~6-fold), but the apparent Ca²⁺ affinity was slightly lower in myocytes for both GCaMP2.2-FKBP12.6 and GCaMP2.2 than in vitro (K_d=1.2 and 1.3 μmol/L, respectively). GCaMP2.2Low and GCaMP2.2Low-FKBP also exhibited roughly a two-fold decrease in apparent affinity in the myocyte environment (K_d~11 μmol/L). While the GCaMP2.2-FKBP12.6 could increase Ca²⁺ buffering in the cleft, that effect is expected to be small and have limited impact for two reasons. First, there would be at most one GCaMP per RyR, and this is minor compared to the intense >100:1 buffering in the cleft environment.⁶ Endogenous Ca²⁺ binding sites include calmodulin and divalent binding sites on RyR, L-type Ca²⁺ channels and other cleft proteins, membrane sites and ATP. Second, buffering in the cleft will not alter the steady state [Ca²⁺]_cleft at rest, and could only slow achievement of steady state by 1-5 ms during channel opening or closing.⁶

After characterizing our junctional cleft-targeted Ca²⁺ sensors, we expressed them in intact rat cardiac myocytes by adenoviral infection. Myocytes were used for experiments 20-24 hours later. Membrane staining with Di-8-ANNEPS indicated that myocytes retain the T-tubules at which most of the junctions reside, under the culture conditions (Online Fig. III). Moreover, a stringent test for normal junctional function utilized transverse line scan images using Fluo-4 (where spatial drop-out or delayed central activation could indicate altered junctional coupling). The results demonstrated that normal transverse synchrony of SR Ca²⁺ release was well-preserved in myocytes expressing the GCaMP sensors after 24 hr culture (Online Fig. IV). There was no decrement in either the uniformity of local transverse release activation or rate of rise of the Ca²⁺ transients. Thus, the clefts where the GCaMP2.2-FKBP12.6 sensors are targeted function normally during ECC.

FKBP12.6-tagged sensors express in a striated pattern, with intensity maxima spaced ~2 μm apart (Fig. 2A; full images are shown in Online Fig. V). This strongly supports the conclusion that the FKBP12.6-tagged sensors are targeted to RyR at the z-line, as in permeabilized cells. In contrast, the un-tagged sensors show a rather uniform cytosolic distribution (Fig. 2A). To test the Ca²⁺ responsiveness of the new sensors when expressed in intact cells, we first depleted myocytes of Ca²⁺ (using ionomycin and Ca²⁺-free solution), to determine the minimum sensor fluorescence (F_min) (Fig. 2Bb vs a). Then, myocytes were Ca²⁺-overloaded by restoring extracellular Ca²⁺ in Na⁺-free solution to assess maximum fluorescence (F_max) just as hypercontracture begins (Fig. 2Bc). Importantly, GCaMP2.2(Low) and GCaMP2.2(Low)-FKBP12.6 have a similar dynamic range, F_max/F_min, in myocytes (5.5±0.6, n=7 for GCaMP2.2 and 5.7±0.5, n=9 for GCaMP2.2-FKBP12.6). These numbers agree well with those in Fig 1E. Thus, our novel FKBP12.6-tagged Ca²⁺ indicators are highly sensitive to Ca²⁺ and are positioned to measure local cleft [Ca²⁺]_Cleft vs. un-tagged sensors reporting bulk cytosolic [Ca²⁺]_Bulk in intact myocytes.

Figure 2. Adenovirally-expressed GCaMP2.2(Low)-FKBP12.6 and untagged GCaMP2.2(Low) are positioned to measure local cleft [Ca²⁺]_Cleft and, respectively, bulk cytosolic [Ca²⁺]_Bulk, and are highly sensitive to Ca²⁺ in intact myocytes.

(A) Example of myocytes infected with adenoviruses expressing untagged GCaMP2.2Low (Left) and GCaMP2.2Low-FKBP12.6 (Right). Bottom panel shows the fluorescence profile along the longitudinal axis of the myocyte for the area shown in the white square in the top panels. Right and Left panels show limited areas from each myocyte; entire myocyte images are in Online Fig. V. (B) Ca²⁺-sensitivity of GCaMP2.2Low-FKBP12.6 in intact cells. A resting myocyte expressing GCaMP2.2Low-FKBP12.6 was imaged first under control conditions (Ba). The cell was then exposed to Ca²⁺-free external solution in the presence of 10 μmol/L ionomycin (Bb). This resulted in a significant fluorescence drop (F_min). Then, the cell was Ca²⁺-overloaded in a 0Na⁺/1 mmol/L Ca²⁺ external solution, which favors Ca²⁺ entry via both ionomycin and reverse mode Na⁺/Ca²⁺ exchange, to saturate the sensor and determine F_max (Bc). The F_max/F_min ratio for this cell is 6.2.

Larger and faster local Ca²⁺ transients in the junctional space vs. bulk cytosol

We expressed the low-affinity untagged and FKBP12.6-tagged sensors to measure Ca²⁺ transients in the bulk cytosol and junctional cleft, respectively (Fig. 3). Myocytes were field-stimulated at 0.5 Hz and Ca²⁺ transients were measured in linescan mode using a laser scanning confocal microscope. Fig. 3A shows representative Ca²⁺ transients reported by GCaMP2.2Low and GCaMP2.2Low-FKBP12.6. Junctional cleft transients recorded by GCaMP2.2Low-FKBP12.6 are ~2-fold larger (Fig. 3B) and have a much faster upstroke (time-to-peak=46±5 vs. 90±7 ms; Fig. 3C) and decay (decay τ=254±19 vs. 421±42 ms; Fig. 3D) compared to global Ca²⁺ transients measured by the un-targeted GCaMP2.2Low. This is encouraging (see Discussion for inferred [Ca²⁺] values), but based on theoretical models¹¹ and indirect cleft Ca²⁺ transient assessments^12,24 one would expect a much larger difference in both amplitude and kinetics between the targeted and untargeted sensor. An obvious explanation is that the kinetics of local [Ca²⁺] change in the cleft are likely to be very fast, and GCaMP sensors lack the temporal resolution to detect such rapid kinetics. The dissociation rate constant for GCaMP2.2Low (~6 s^-1: Online Table I) is consistent with this notion. While this new low-affinity Ca²⁺-sensor is not fast enough to accurately report [Ca²⁺]_Cleft dynamics during ECC, it may still be useful in determining relative changes in [Ca²⁺]_cleft with acute treatments (e.g. slower or larger). Targeted GCaMP2.2Low may also provide more accurate values of local [Ca²⁺]_i in microdomains with local Ca²⁺ flux rates that are lower than in striated muscle junctions, where local Ca²⁺ flux rates may be among the highest in nature.

Figure 3. Electrically stimulated Ca²⁺ transients measured with GCaMP2.2Low and GCaMP2.2Low-FKBP12.6.

(A) Representative recordings in myocytes expressing GCaMP2.2Low and GCaMP2.2Low-FKBP12.6. (B-D) Amplitude (B), time-to-peak (C) and decay time (D) of Ca²⁺ transients reported by GCaMP2.2Low (GCaMP) and GCaMP2.2Low-FKBP12.6 (FKBP). Data are the mean±SEM of 10 (GCaMP2.2Low) and 13 (GCaMP2.2Low-FKBP12.6) separate experiments.

Upon β-adrenergic activation with isoproterenol (ISO; Online Fig VI), untargeted GCaMP2.2Low readily detects the expected large increase in Ca²⁺ transient amplitude and two-fold acceleration of [Ca²⁺]_i decline due to enhanced SR Ca²⁺ uptake. The cleft-targeted sensor Ca²⁺ transient decays much faster than the untargeted sensor under control conditions, but fails to accelerate with ISO, despite a similar doubling of amplitude. Since [Ca²⁺]_cleft is expected to sense released Ca²⁺ and diffusion from the cleft to cytosol, rather than SR Ca²⁺ uptake rate (which is distributed throughout the cytosol), this is exactly what one would expect. Thus, the cleft-targeted GCaMP2.2Low is more sensitive to Ca²⁺ release at the cleft than to transport by SERCA, and differs qualitatively from bulk [Ca²⁺]_i sensors. This relates to Ca²⁺ sparks which decay primarily via diffusion from the source, with minor influence by SERCA function,³² despite much of the indicator signal coming from outside the cleft. In contrast, global [Ca²⁺]_i decline is dictated almost entirely by Ca²⁺ transport out of the cytosol.⁶

Standing diastolic [Ca²⁺] gradient between the junctional space and bulk cytosol

During diastole, Ca²⁺ leak from the SR or Ca²⁺ influx across sarcolemma may raise local [Ca²⁺]_Cleft above [Ca²⁺]_Bulk (distant from Ca²⁺ sources). Moreover, the very high density of both L-type Ca²⁺ channels and RyR2 at the cleft, compared to the broader distribution of the transporters that remove Ca²⁺ from the cytosol (SR Ca²⁺-ATPase and Na⁺/Ca²⁺ exchange) could allow even modest Ca²⁺ leak through these channels to produce a standing [Ca²⁺] gradient (Fig 4A, left). Increases in SR Ca²⁺ leak have been implicated in pathological states, where [Ca²⁺]_Cleft could be preferentially elevated. We used our targeted GCaMP2.2 to directly test for such diastolic standing [Ca²⁺] gradients between the cleft and the cytosol.

Figure 4. Measurement of standing diastolic Ca²⁺ gradient between cleft and bulk cytosol.

(A) Cartoon illustrating the expected effect of blocking Ca²⁺ fluxes through RyRs, NCX and L-type Ca²⁺ channels on [Ca²⁺]_Cleft and [Ca²⁺]_Bulk. We hypothesize that [Ca²⁺]_Cleft > [Ca²⁺]_Bulk during diastole (left). Upon inhibition of both SR Ca²⁺ leak and sarcolemmal Ca²⁺ fluxes, [Ca²⁺] decreases in both compartments and at steady-state [Ca²⁺]_Cleft = [Ca²⁺]_Bulk (right). (B-C) Representative examples of the change in the fluorescence of GCaMP2.2-FKBP12.6 (B) and GCaMP2.2 (C) during application and washout of 1 mmol/L tetracaine and 0Na⁺/0Ca²⁺ external solution. Myocytes were first field-stimulated (at 0.5 Hz) to reach steady-state, then stimulation was stopped and we added tetracaine and 0Na⁺/0Ca²⁺ solution. Top panels show the fluorescence 2D images of myocytes expressing GCaMP2.2-FKBP12.6 (B) and GCaMP2.2 (C) at the time points (a-c) indicated in the lower panels. Lower panels show time-dependence of average fluorescence. Fluorescence of both sensors decreased rapidly in the presence of tetracaine and 0Na⁺/0Ca²⁺ solution and the decrease was larger for GCaMP2.2-FKBP12.6. (D) Mean decrease in fluorescence intensity upon application of tetracaine and 0Na⁺/0Ca²⁺ external solution for myocytes expressing GCaMP2.2-FKBP12.6 and untagged GCaMP2.2 (n=6 cells for each). (E) Conversion of data in panel (D) to [Ca²⁺].

Upon blocking Ca²⁺ fluxes into the cleft, [Ca²⁺] should decline in both compartments, and importantly, any [Ca²⁺]_Cleft vs. [Ca²⁺]_Bulk gradient should dissipate (Fig. 4A, right). This maneuver is valuable so that we can define a point (F₀) where both the cytosolic and targeted GCaMP2.2 sense the same [Ca²⁺]_i. Figure 4B-C shows how [Ca²⁺]_Cleft and [Ca²⁺]_Bulk decline when we abruptly block RyRs with 1 mM tetracaine and both LTCC and NCX by rapidly switching the external solution to a 0Na⁺/0Ca²⁺ solution. Both [Ca²⁺]_Cleft and [Ca²⁺]_Bulk dropped rapidly after the application of tetracaine and 0Na⁺/0Ca²⁺ solution, reaching the F₀ steady-state (where [Ca²⁺]_Cleft =[Ca²⁺]_Bulk) in about 1 min (Figs. 4B-C, transition from time a to b). Note that [Ca²⁺] decline was much larger for GCaMP2.2-FKBP12.6. These results suggest that the initial diastolic [Ca²⁺]_Cleft was higher (F/F₀ =1.8) than [Ca²⁺]_Bulk (F/F₀ =1.2). In both cases, fluorescence was restored on return to the normal bath solution (time point c in Figs. 4B-C).

Mean data confirm that inhibition of SR and sarcolemmal Ca²⁺ fluxes in the cleft produces a significantly larger decline in the fluorescence of GCaMP2.2-FKBP12.6, which reports local [Ca²⁺]_Cleft, vs. the untagged GCaMP2.2, which reports [Ca²⁺]_Bulk (-ΔF/F₀=0.55±0.10 for GCaMP2.2-FKBP12.6 vs. 0.13±0.02 for GCaMP2.2; Fig. 4D). This indicates that there is indeed a standing [Ca²⁺]_i gradient during diastole between the junctional cleft and bulk cytosol. Assuming that diastolic [Ca²⁺]_Bulk =100 nmol/L (prior to tetracaine, 0Na⁺/0Ca²⁺) and using data in Fig. 1E, we calculated that diastolic [Ca²⁺]_Cleft is 194±28 nmol/L.

Another way of reducing diastolic SR Ca²⁺ leak is to block SERCA with thapsigargin (Fig 5A). SERCA inhibition slowly unloads the SR, which is paralleled by a slow decrease in SR Ca²⁺ leak.³³ As expected, with 10 μM thapsigargin and 0Na⁺/0Ca²⁺ solution, the fluorescence signal in myocytes expressing either GCaMP2.2 or GCaMP2.2-FKBP12.6 declined more slowly, reaching a new steady-state in ~10 min (Fig. 5B). A similar time course was previously demonstrated for thapsigargin-induced cessation of SR Ca²⁺ leak, measured as the rate of decline in SR [Ca²⁺].^33,34 Similar to RyR blockade with tetracaine, the fluorescence decline was significantly larger for GCaMP2.2-FKBP12.6 vs. untagged GCaMP2.2 (-ΔF/F₀=0.77±0.11 vs. 0.41±0.06; Fig. 5C). The overall decline in [Ca²⁺]_i was larger for thapsigargin vs. tetracaine (presumably due to longer time in Ca²⁺-free solution) but the difference between cleft and bulk was similar for both methods (-ΔF/F₀=0.36 vs. 0.42).

Figure 5. The standing diastolic [Ca²⁺]_i gradient is caused mainly by the SR Ca²⁺ leak.

(A) Cartoon illustrating a second way of progressive reduction of diastolic SR Ca²⁺ leak by blocking SERCA with 10 μmol/L thapsigargin. (B) Time course of GCaMP2.2-FKBP12.6 and GCaMP2.2 fluorescence during application of thapsigargin and 0Na⁺/0Ca²⁺ external solution (mean of 8 separate experiments). (C) Mean decrease in fluorescence intensity in myocytes expressing GCaMP2.2-FKBP12.6 and untagged GCaMP2.2 upon inhibition of both SR and sarcolemmal Ca²⁺ fluxes with thapsigargin and 0Na⁺/0Ca²⁺ external solution (SR Leak+SL Influx), blockade of sarcolemmal Ca²⁺ transport in 0Na⁺/0Ca²⁺ solution (SL Influx) and inhibition of SR Ca²⁺ leak by emptying the SR with thapsigargin (SR Leak).

To assess the relative contribution of the SR and sarcolemmal Ca²⁺ fluxes to Ca²⁺ entry into the cleft, we did separate experiments where only sarcolemmal Ca²⁺ fluxes are blocked (0Na⁺/0Ca²⁺ solution). In this case, the decrease in the sensor fluorescence was similar for the FKBP12.6-tagged and the untagged sensor (-ΔF/F₀=0.22±0.03 for GCaMP2.2-FKBP12.6 vs. 0.17±0.02 for GCaMP2.2; Fig. 5C). In contrast, inhibition of SR Ca²⁺ leak alone resulted in a significantly larger drop in the GCaMP2.2-FKBP12.6 fluorescence vs. the untagged sensor (-ΔF/F₀=0.55±0.11 vs. 0.24±0.07; Fig. 5C). These results indicate that most of the Ca²⁺ entering the junctional cleft during diastole comes from the SR and SR Ca²⁺ leak is the main factor in setting a diastolic [Ca²⁺]_i gradient between the cleft and bulk cytosol.

DISCUSSION

In many cell types, [Ca²⁺]_i is regulated and signals differently in various subcellular microdomains, which greatly enhances its versatility as a secondary messenger. Recognition of such subcellular regulation stimulated us to develop methods for measuring local [Ca²⁺]_i in microdomains. Here, we attached the genetically encoded Ca²⁺ sensor GCaMP2.2 and a newly developed lower Ca²⁺ affinity variant GCaMP2.2Low to the N-terminus of FKBP12.6 to construct novel Ca²⁺ sensors that bind specifically to RyRs. Our data indicate that GCaMP2.2Low-FKBP12.6 has a high affinity and selectivity for RyRs, although slightly reduced compared to untagged FKBP12.6. In cardiac myocytes, RyRs are localized predominantly in the junctional SR membrane, where the SR comes into close proximity of the external sarcolemma (the dyads). Due to this specific targeting, the FBP12.6-tagged sensors report local [Ca²⁺]_i in the small dyadic cleft, [Ca²⁺]_Cleft. This local [Ca²⁺]_Cleft controls and is influenced by both RyRs and L-type Ca²⁺ channels. Therefore, [Ca²⁺]_Cleft critically affects myocyte function at rest (during diastole) and during electrical excitation. Despite its importance, methods to directly measure [Ca²⁺]_Cleft are only now becoming available.^21,22 This key parameter has typically been inferred from the dynamics of NCX^12,24 and L-type Ca²⁺ currents^12,23 in electrically stimulated myocytes, but no such estimates could be made for diastolic [Ca²⁺]_Cleft. Our novel Ca²⁺ sensors allow direct measurement of [Ca²⁺]_Cleft, both during ECC (the low affinity sensor GCaMP2.2Low-FKBP12.6) and at rest (GCaMP2.2-FKBP12.6). However, the extremely rapid changes of [Ca²⁺]_Cleft expected during ECC are too fast to be accurately captured by even the 10 μM K_d sensor, although they are likely to be useful to assess relative changes in [Ca²⁺]_cleft during perturbations that might alter the amplitude or kinetics of [Ca]_cleft changes during ECC (as we have seen for ISO stimulation).

We use FKBP12.6 to target the sensors because of the very high affinity and selectivity for RyR2 in cardiac myocytes.³¹ A limitation with this approach is that FKBP12.6 can also influence RyR channel gating, stabilizing the closed state.³⁵ FKBP12.6 overexpression in myocytes can reduce spontaneous Ca²⁺ spark frequency and either increase or decrease Ca²⁺ transient amplitude.^36-37 Thus, using FKBP12.6 to target these Ca²⁺ sensors may reduce diastolic RyR2 opening. On the other hand, in studies measuring simultaneously FKBP12.6 binding to RyR2 and Ca²⁺ sparks, saturating RyR2 with FKBP12.6 only reduced Ca²⁺ spark frequency by 18%.³¹ This effect would mean that we are probably slightly underestimating diastolic [Ca²⁺]_Cleft.

A fraction (~15-20%) of RyRs are located outside the sarcolemma-SR junctions in rat myocytes.³⁸ Thus, the GCaMP2.2-FKBP12.6 signal is not generated entirely by [Ca²⁺]_Cleft, but is slightly contaminated by [Ca²⁺] at such non-junctional RyRs. Of note, local [Ca²⁺] at non-junctional RyRs may also be relevant, even if less influential on sarcolemmal currents. The adenoviral expression of GCaMP sensors in cardiac myocytes could risk alteration of dyad structure or function. While cleft dimensions were not directly measured (e.g. via electron microscopy), the FKBP12.6-targeted sensors are highly localized along T-tubules (of which ~50% are in dyads in rat myocytes)⁶ and normal synchronous transverse activation during ECC was maintained (Online Fig. IV). Therefore, we assume that the ultrastructure of the junctional cleft was not appreciably altered by the experimental conditions. FKBP12.6-GCaMP2.2 also has to be studied in a practical range of protein expression. If one waits too long or transfects too aggressively, one can express enough FKBP12.6-targeted protein to more than saturate the ~1 μM of RyR binding sites. In that case, the excess sensor will not be effectively targeted. Any amount of untargeted GCaMP2.2-FKBP12.6 (as above) would cause underestimation of the difference between [Ca²⁺]_Cleft and [Ca²⁺]_Bulk. This means that our [Ca²⁺]_Cleft measurements and the diastolic [Ca²⁺] gradient are lower limit estimates. True [Ca²⁺]_Cleft would be larger. On balance, with awareness of these challenges with respect to interpretation, we think the advantages of targeting sensors to the cleft this way outweigh the limitations, and do not affect any of our conclusions. Potential enhancements in live cell imaging resolution that selectively visualize such local domains would complement approaches like ours with locally target sensors.²²

[Ca²⁺]_Cleft and [Ca²⁺]_Bulk in cardiac myocytes

Our novel cleft-targeted sensor GCaMP2.2Low-FKBP12.6 reported a larger and faster rise in [Ca²⁺]_Cleft vs. [Ca²⁺]_Bulk during electrical excitation of myocytes, in good agreement with previous estimates from indirect measurements.^9-12 Weber et al.²⁴ inferred the submembrane [Ca²⁺] near NCX during a normal action potential as peaking at >3.2 μM within <32 ms (vs. a global Ca²⁺ transient that peaks at 1.1 μM in 81 ms). However [Ca²⁺]_Cleft ought to be much higher than submembrane [Ca²⁺]_i. Acsai et al.¹² estimated the Ca²⁺ transient near Ca²⁺ release sites from the kinetics of L-type Ca²⁺ current and from a fraction of NCX current. They calculated that this local Ca²⁺ transient has 10-15 μM in amplitude, peaks at about 10 ms after depolarization and recovers within ~50 ms. In contrast, the global Ca²⁺ transient reached a maximum of ~1 μM within ~70-100 ms. Estimates of [Ca²⁺]_Cleft based on geometry, diffusion and Ca²⁺ flux measurements suggest even higher (100 μM) and faster [Ca²⁺]_Cleft changes during electrical excitation.¹¹ Based on the measured diastolic [Ca²⁺]_cleft and [Ca²⁺]_bulk (194 nmol/L and 100 nmol/L, respectively, Fig. 4E), the characteristics of GCaMP2.2Low and GCaMP2.2Low-FKBP response to Ca²⁺ in myocytes (Fig. 1E), and the Ca²⁺ transient amplitude (Fig. 3C), we calculated that peak [Ca²⁺]_bulk=560±73 nmol/L while peak [Ca²⁺]_cleft is ~2-fold larger (1272±199 nmol/L). This peak [Ca²⁺]_cleft is much smaller than previous estimates. The time-to-peak of [Ca²⁺]_Cleft was 46±5 ms (vs. 90±7 ms for [Ca²⁺]_Bulk), slightly longer than the estimates based on NCX and L-type Ca²⁺ current. We believe that these differences in the amplitude and dynamics of cleft Ca²⁺ transients vs. previous estimates are mostly due to the relatively slow kinetics of Ca²⁺ binding and unbinding of the GCaMP2.2 sensors (see Online Table I). Thus, while GCaMP2.2Low detects larger and faster cleft Ca²⁺ transients vs. the global sensor and can detect changes in amplitude and kinetics (e.g. with ISO), much faster indicators will be required for accurate [Ca²⁺]_Cleft measurement during the dynamics of active SR Ca²⁺ release in heart.

Using GCaMP2.2-FKBP12.6 we could directly assess diastolic [Ca²⁺]_Cleft, a parameter not previously measured. We demonstrated that during diastole, [Ca²⁺]_Cleft is >90 nmol/L higher than [Ca²⁺]_Bulk. This standing diastolic [Ca²⁺] gradient indicates that Ca²⁺ extrusion from the cleft (via junctionally-located NCX and diffusion into bulk cytosol) cannot keep up with Ca²⁺ entry into the cleft. Our data show that diastolic cleft Ca²⁺ influx is caused mainly by the SR Ca²⁺ leak (vs. entry through LTCC). At the same time, elevated diastolic [Ca²⁺]_Cleft can reduce LTCC availability (via Ca²⁺ -dependent inactivation) and alter RyR gating to increase their activity by direct RyR activation, and also indirectly via activation of CaMKII followed by CaMKII-dependent RyR phosphorylation.³⁹ Increased RyR opening leads to a larger SR Ca²⁺ leak and a higher risk for generation of propagating Ca²⁺ waves, which are a known risk for arrhythmias. The balance of Ca²⁺ fluxes in the cleft is altered under pathological conditions, e.g. in heart failure SR Ca²⁺ leak and Na⁺/Ca²⁺ exchange function are increased.⁴⁰

The high diastolic [Ca²⁺]_Cleft may also have important effects on local signaling cascades beyond the sphere of E-C coupling. Indeed, there is much indirect evidence that local [Ca²⁺]_i in these environments may be important in the activation of calcineurin or CaMKII that can signal to the nucleus and trigger alterations in gene expression.⁴¹ In pathological conditions where SR Ca²⁺ leak is known to be enhanced, the higher local level of diastolic [Ca²⁺]_Cleft may have an important impact on these (and other) broader signaling pathways.

In summary, we have developed sensors for measuring [Ca²⁺]_Cleft vs. [Ca²⁺]_Bulk and demonstrated dynamic differences during excitation-contraction coupling. We have also provided measures of a normal standing [Ca²⁺]_i gradient during diastole between the junctional cleft and the bulk cytosol and showed that this gradient is set mainly by the SR Ca²⁺ leak.

Novelty and Significance.

What Is Known?

• Ca²⁺ is regulated and signals differently in various subcellular microdomains, which greatly enhances its second messenger versatility.

• In the heart, sarcoplasmic reticulum (SR) Ca²⁺ release is controlled by local [Ca²⁺]_i in the junctional cleft, the small space between sarcolemma and junctional SR.

• This local cleft [Ca²⁺]_i is thought to be important in both the regulation of excitation-contraction coupling, but also to downstream Ca²⁺-dependent signaling (e.g. via calmodulin, calcineurin and CaMKII).

What New Information Does This Article Contribute?

• We constructed novel Ca²⁺-sensors that are targeted to microdomains rich in ryanodine receptors.

• In cardiac myocytes, the novel Ca²⁺-sensors are positioned to measure local Ca²⁺ in the junctional cleft.

• There is a standing Ca²⁺ gradient during diastole between the junctional cleft and bulk cytosol in cardiac myocytes, which could influence local Ca²⁺-dependent signaling within the cleft.

Ca²⁺ is a universal second messenger involved in activation/regulation of very diverse cellular processes. To allow such versatility, Ca²⁺ is regulated and signals differently in various subcellular microdomains, which generated a large interest in developing methods for measuring local rather than bulk Ca²⁺. Here we targeted novel Ca²⁺-sensors to microdomains rich in ryanodine receptors. This was accomplished by fusing the genetically encoded Ca²⁺-sensor GCaMP2.2 and a lower Ca²⁺-affinity variant GCaMP2.2Low to FKBP12.6, a protein that binds with high affinity and selectivity to ryanodine receptors. In cardiac myocytes, these sensors report local [Ca²⁺] in the small junctional cleft between the plasmalemma and junctional sarcoplasmic reticulum ([Ca²⁺]_Cleft), a parameter that critically affects heart function and dysfunction. Using these sensors, we demonstrate that [Ca]_cleft reaches higher levels with faster kinetics than global [Ca²⁺]_i during excitation-contraction coupling. We also show that there is a substantial standing diastolic [Ca²⁺] gradient between [Ca²⁺]_Cleft and bulk [Ca²⁺]_i, due mainly to ryanodine receptor-dependent Ca²⁺ leak. Such [Ca²⁺]_Cleft characteristics may have important impact on local Ca²⁺-dependent signaling.

Nonstandard Abbreviations and Acronyms

[Ca²⁺]_i

intracellular Ca²⁺ concentration

[Ca²⁺]_Cleft

local [Ca²⁺] in the junctional cleft

[Ca²⁺]_Bulk

[Ca²⁺] in the bulk cytosol

CaM

calmodulin

ECC

excitation – contraction coupling

GECI

genetically-encoded Ca²⁺ indicator

LTCC

L-type Ca²⁺ channels

NCX

Na⁺/Ca²⁺ exchanger

RyR

ryanodine receptors

SR

sarcoplasmic reticulum