Local recovery of cardiac calcium‐induced calcium release interrogated by ultra‐effective, two‐photon uncaging of calcium

Abstract

Key points

• In cardiac myocytes, subcellular local calcium release signals, calcium sparks, are recruited to form each cellular calcium transient and activate the contractile machinery.

• Abnormal timing of recovery of sparks after their termination may contribute to arrhythmias.

• We developed a method to interrogate recovery of calcium spark trigger probabilities and their amplitude over time using two‐photon photolysis of a new ultra‐effective caged calcium compound.

• The findings confirm the utility of the technique to define an elevated sensitivity of the calcium release mechanism in situ and to follow hastened recovery of spark trigger probabilities in a mouse model of an inherited cardiac arrhythmia, which was used for validation.

• Analogous methods are likely to be applicable to investigate other microscopic subcellular signalling systems in a variety of cell types.

Abstract

In cardiac myocytes Ca²⁺‐induced Ca²⁺ release (CICR) from the sarcoplasmic reticulum (SR) through ryanodine receptors (RyRs) governs activation of contraction. Ca²⁺ release occurs via subcellular Ca²⁺ signalling events, Ca²⁺ sparks. Local recovery of Ca²⁺ release depends on both SR refilling and restoration of Ca²⁺ sensitivity of the RyRs. We used two‐photon (2P) photolysis of the ultra‐effective caged Ca²⁺ compound BIST‐2EGTA and laser‐scanning confocal Ca²⁺ imaging to probe refractoriness of local Ca²⁺ release in control conditions and in the presence of cAMP or low‐dose caffeine (to stimulate CICR) or cyclopiazonic acid (CPA; to slow SR refilling). Permeabilized cardiomyocytes were loaded with BIST‐2EGTA and rhod‐2. Pairs of short 2P photolytic pulses (1 ms, 810 nm) were applied with different intervals to test Ca²⁺ release amplitude recovery and trigger probability for the second spark in a pair. Photolytic and biological events were distinguished by classification with a self‐learning support vector machine (SVM) algorithm. In permeabilized myocytes data recorded in the presence of CPA showed a lower probability of triggering a second spark compared to control or cAMP conditions. Cardiomyocytes from a mouse model harbouring the arrhythmogenic RyR^R420Q mutation were used for further validation and revealed a higher Ca²⁺ sensitivity of CICR. This new 2P approach provides composite information of Ca²⁺ release amplitude and trigger probability recovery reflecting both SR refilling and restoration of CICR and RyR Ca²⁺ sensitivity. It can be used to measure the kinetics of local CICR recovery, alterations of which may be related to premature heart beats and arrhythmias.

Key points

• In cardiac myocytes, subcellular local calcium release signals, calcium sparks, are recruited to form each cellular calcium transient and activate the contractile machinery.

• Abnormal timing of recovery of sparks after their termination may contribute to arrhythmias.

• We developed a method to interrogate recovery of calcium spark trigger probabilities and their amplitude over time using two‐photon photolysis of a new ultra‐effective caged calcium compound.

• The findings confirm the utility of the technique to define an elevated sensitivity of the calcium release mechanism in situ and to follow hastened recovery of spark trigger probabilities in a mouse model of an inherited cardiac arrhythmia, which was used for validation.

• Analogous methods are likely to be applicable to investigate other microscopic subcellular signalling systems in a variety of cell types.

Introduction

Cardiac muscle contractions are controlled by transient elevations of the intracellular Ca²⁺ concentration in each cardiomyocyte. A large fraction of these transients is governed by Ca²⁺ released from the sarcoplasmic reticulum (SR) via Ca²⁺ release channels located on the SR membrane. These channels are ryanodine receptors (RyRs, RyR2 in the heart) and are activated by the Ca²⁺‐induced Ca²⁺ release (CICR) mechanism by Ca²⁺ entering via L‐type Ca²⁺ channels (for review see Bers, ²⁰⁰²). SR Ca²⁺ release occurs via summation of elementary microscopic Ca²⁺ signalling events, termed Ca²⁺ sparks. While usually triggered by Ca²⁺ entry, sparks can also occur spontaneously between heartbeats.

The RyRs form tetrameric macromolecular complexes (2.2 MDa) which contain several accessory proteins (e.g. protein kinases and phosphatases), and their Ca²⁺ signalling function is regulated by multiple post‐translational protein modifications, including phosphorylation of several sites, but also by oxidation (for review see Niggli et al. ²⁰¹³). More than 200 mutations of the RyR2 have been identified, most of them resulting in a gain of function with abnormally high Ca²⁺ sensitivity. This can lead to untimely (i.e. diastolic) larger spontaneous Ca²⁺ release events, such as Ca²⁺ waves. The latter have an arrhythmogenic potential, by causing delayed afterdepolarizations. Consequently, many patients harbouring RyR2 mutations suffer from potentially life threatening catecholaminergic polymorphic ventricular tachycardias (CPVTs), often precipitated by emotional stress or physical activity.

A variety of approaches have been developed to examine the function of the RyRs from the molecular to the cellular level. For example, RyRs can be incorporated into artificial lipid bilayers to perform single channel electrophysiological studies (e.g. Cannell et al. ²⁰¹³), while purified SR vesicles containing RyRs are amenable to a variety of biochemical functional experiments (Ferrero et al. ²⁰⁰⁷). This allows examination of ion binding and transport, and even estimation of RyR open probability, because ryanodine only binds to the open channels. In isolated cells, information about the RyRs and CICR can be obtained by recording transient whole‐cell Ca²⁺ signals, but also by evaluating spontaneous or triggered Ca²⁺ sparks and waves. While the biochemical studies using SR vesicles usually examine the RyR function with a low temporal resolution or in the steady‐state, the cellular Ca²⁺ signalling events are rapid and occur in the millisecond time domain, representing a challenge for most imaging techniques.

Analysis of spontaneous Ca²⁺ sparks provides information about RyR function at a near molecular level, while the channels are still in their normal environment within the Ca²⁺ release unit (CRU). This approach is a powerful way to understand functional alterations of the RyRs resulting, for example, from post‐translational modifications or from CPVT and other RyR mutations. Unfortunately, the experimenter has very limited control over the occurrence of spontaneous Ca²⁺ sparks, and if the spark frequency is low, a reasonable analysis, which requires a substantial number of events to be evaluated, is not feasible.

Methods to overcome this experimental limitation have been developed in the past. One technique makes use of long‐lasting RyR openings to a subconductance state that can be elicited by a very low dose of ryanodine (Ramay et al. ²⁰¹¹). Thereby, one channel within a RyR cluster is thought to drive repetitive Ca²⁺ sparks originating from that particular cluster. Analysis of amplitude recovery and the distribution of the intervals can be performed to extract information about the RyRs. However, this method relies on a distinctly non‐physiological stimulus regime to elicit Ca²⁺ release from the SR.

Here we present an alternative solution. Making use of the new caged Ca²⁺ compound BIST‐2‐EGTA (Agarwal et al. ²⁰¹⁶), protocols with paired two‐photon (2P) photolytic trigger pulses can be applied to elicit Ca²⁺ sparks at short to long intervals. This new caged molecule is vastly superior to previously available compounds. Importantly, the two‐photon absorbance cross section is 350 Goeppert‐Mayer (GM), while for DM‐nitrophen it is only 0.01 GM. This allows much lower concentrations of the compound to be used, much shorter illumination times to be applied and requires dramatically reduced 2P photolytic power levels. Because of these superior features of the new compound, such paired 2P protocols can be applied, for the first time, with physiologically relevant timing (i.e. in the millisecond range). Since in confocal line scan recordings these photolytic Ca²⁺ signals have spatiotemporal features that are rather comparable to the biological Ca²⁺ sparks themselves, a machine learning approach was adopted to distinguish between the two similar events during the image analysis. This involved the use of support vector machines (SVMs), supervised learning models, which analyse data and recognize patterns. They are frequently used for classifications tasks (Noble, 2006). In this paper, we introduce this procedure and validate the technique by using pharmacological modifiers for CICR and a transgenic mouse model harbouring the arrhythmogenic RyR^R420Q mutation (Domingo et al. ²⁰¹⁵; Yin et al. ²⁰²¹).

Methods

Ethical approval

Housing and breeding of the mice was carried out in the departmental animal facility, with free access to rodent laboratory chow and water and a 12:12 h day–night cycle. Handling was performed with permission of the State Veterinary Administration of the Canton of Bern (BE) and in accordance with Swiss Federal Animal protection law. The animal experimentation permit (No. BE 6/2019) was issued after ethical review of our experimental and preparatory procedures, including the death of the mice (see below), by the State Committee for Animal Ethics and with endorsement by the State Veterinarian of the Canton of Bern, and after final approval by the Swiss Federal Food Safety and Veterinary Office. All experiments were carried out in accordance with the guidelines and regulations on animal experimentation of The Journal of Physiology (Grundy, 2015).

Animals and cardiac myocytes isolation

Ventricular myocytes were isolated from C57Bl/6 or homozygous RyR2^R420Q knock‐in male mice, aged between 3 and 11 months (Wang et al. ²⁰¹⁷) following an established protocol (Louch et al. ²⁰¹¹). Adult mice were euthanized by cervical dislocation and the hearts were excised, cannulated and retrogradely perfused on a Langendorff system. Hearts were perfused at 37°C for around 15 min with a Ca²⁺‐free modified Tyrode solution composed of (mmol/l): 140 NaCl, 5.4 KCl, 1.1 MgCl₂, 10 HEPES, 1 NaH₂PO₄, 10 glucose (pH 7.4). Cells were enzymatically dissociated using a cocktail of collagenase type II (160 U/ml, Allschwil, Worthington, Switzerland) and protease type XIV (0.21 U/ml, Buchs, Sigma, Switzerland). After isolation, ventricular myocytes were kept at room temperature in a modified Tyrode solution containing 250 μmol/l CaCl₂ and used within 6 h.

Two‐photon photolysis in permeabilized cardiac myocytes

Isolated ventricular myocytes were permeabilized by exposure to β‐escin for 60 s (in mmol/l: 100 potassium aspartate, 20 KCl, 3.7 MgCl₂, 0.5 EGTA, 10 HEPES, 0.005% β‐escin; pH 7.2). Permeabilized myocytes were then placed in a recording chamber in a final solution containing (mmol/l): 120 potassium aspartate, 3 K₂ATP, 0.1 EGTA, 3 MgCl₂, 10 phosphocreatine, 5 U/ml creatine phosphokinase, 10 HEPES, 1 l‐glutathione reduced, 0.04 CaCl₂, 0.05 rhod‐2‐K₃, 0.05 BIST‐2EGTA‐K₈; final [Ca²⁺]_i was 60 nmol/l, pH 7.2. Free Ca²⁺ concentration was calculated using Patcher's Power Tools package for Igor Pro (WaveMetrics, Lake Oswego, OR, USA). In pharmacological experiments 5 μmol/l cAMP, 1 μmol/l cyclopiazonic acid (CPA) or 10 mmol/l caffeine together with 20 mmol/l 2,3‐butanedione 2‐monoxime (BDM) was added.

A mode‐locked Ti:sapphire laser was used for 2P excitation of BIST‐2EGTA. The wavelength of the laser was set to 810 nm with pulse duration of ∼120 fs.

The power of the laser was adjusted to 20 mW by neutral density and polarizing filters and was measured at the objective focal plane by a power meter (PM200 with sensor S170C, Thorlabs, Newton, NJ, USA). The laser beam was guided to the second SIM scanner of the confocal microscope (Fluoview 1000, Olympus, Volketswil, Switzerland) operating in single point excitation mode simultaneously with the main scanner.

Pairs of photolytic pulses (Fig. 1A ) were triggered by the confocal microscope using the Time Controller module in Fluoview 1000 software. Intra‐pair interval as well as duration of individual photolytic pulses was controlled by two daisy‐chained electronic shutters LS3 (Vincent Associates, Rochester, NY, USA). The duration of the individual photolytic pulse was 1 ms. The main scanner of the confocal microscope was operating in line scan mode. To record changes in Ca²⁺ concentration rhod‐2 (Biotium Inc., Hayward, CA, USA) was excited at 561 nm and detected at 585–685 nm.

Figure 1. Two photon photolysis, Ca²⁺ event analysis and classification.

A–C, line scan (x–t) confocal images recorded in separate permeabilized cardiac myocytes in final experimental solution in the presence of 10 mmol/l caffeine and 20 mmol/l BDM (A), in control conditions (B) and in the presence of 5 μmol/l cAMP (C). Vertical red lines represent positions of 2P pulses. Below each line scan is the fluorescence time profile calculated by averaging a 1 μm region (black rectangle in respective line scan) around the centres of triggered Ca²⁺ events. The blue lines are fits of events with sum of model spike functions. D, zoom into pairs of triggered Ca²⁺ events. Magenta lines represent detected amplitudes of events. Amplitude ratios (A ₂/A ₁) are calculated from individual pairs of events. E, scatter plot of training dataset (measured photolytic signals together with generated pseudo‐data) visualized using two principal components (PC 1 and PC 2; total variance explained by two principal components was 94.6%). Green line represents decision border of trained OCSVM classifier (σ = 0.005183, ν = 0.002441). F, scatter plot showing classification of all measured events using the trained OCSVM classifier.

All experiments were performed at room temperature (∼22°C). Chemicals were purchased from Sigma‐Aldrich unless stated otherwise.

Statistical analysis

Statistical significance was tested using generalized linear mixed models (Sikkel et al. ²⁰¹⁷) with two levels (isolation and cell) on the whole dataset of calcium events. Pairwise comparison with Holm–Bonferroni correction was used to test significant differences among individual groups. Data are shown as means ± 95% confidence intervals estimated from mixed models or boxplots. The N corresponds to the number of animals, and the n to number of cells. In the figures, horizontal black lines indicate significance; for details see Supporting information, Statistical Summary Document. All image and statistical analyses were performed in MATLAB (R2021a, The MathWorks, Natick, MA, USA).

Results

Local Ca²⁺ events analysis and classification

Two‐photon photolytic excitation allows uncaging of Ca²⁺ from the near diffraction limited volume of the exciting laser only, thereby enabling precise spatial and temporal control over the photochemical signal used to trigger the biological Ca²⁺ sparks. But the photoreleased Ca²⁺ will immediately (within less than 1 ms) spread by diffusion thereby giving rise to an expanded Ca²⁺ signal, similar to the Ca²⁺ signals of biological Ca²⁺ sparks. Therefore, a major challenge was to distinguish between Ca²⁺ released only from BIST‐2EGTA and the actually triggered Ca²⁺ spark, since both signals can be very similar.

For this purpose, line scan (x–t) images of Ca²⁺ fluorescence recorded with rhod‐2 were analysed using a custom program written in MATLAB. A train of light flashes from a mode‐locked Ti:sapphire laser triggered local Ca²⁺ events from the BIST‐2EGTA:Ca complex (Fig. 1A and B). These were detected by an approach developed by Bankhead et al. (2011). First, the fluorescence time profile was calculated by averaging a 1 μm region around the event centres, which were determined from the images filtered by a median filter (kernel (5 x 5)). This time profile was further normalized by fitting the baseline with a high order polynomial function or spline to ΔF/F ₀ units. After this step, individual pairs of triggered Ca²⁺ events were fitted by the sum of model spike functions (Fig. 1B and C) (Zahradníková et al. ²⁰⁰⁷; Janicek et al. ²⁰¹³). Fits of individual Ca²⁺ events were then used to estimate their parameters such as amplitude, time to peak (TTP), full duration at half maximum (FDHM), area under the fit and time constant of exponential decay (τ_decay). Full width at half maximum (FWHM) was estimated from the fluorescence profile, calculated by averaging a 5 ms region around the maximum in the temporal fluorescence profile of the Ca²⁺ event, fitted by a single Gaussian function.

Ca²⁺ events generated in the presence of 10 mmol/l caffeine to empty the SR and 20 mmol/l BDM were considered to be photolytic signals only (Ca²⁺ released only from BIST‐2EGTA, no Ca²⁺ spark triggered) regardless of the mouse strain used (see Fig. 1A and D left panel). These purely photolytic signals still exhibited some variability which may arise from the noise in the line scan images, the kinetics of the Ca²⁺indicator and the limited sampling frequency. Since we had these experimental examples of purely photolytic signals, we decided to describe precisely the distribution of this class (target class) and detect data from another class (biological Ca²⁺ sparks) as outliers. To do so, we trained a one‐class support vector machine (OCSVM, Schölkopf et al. ²⁰⁰¹) with a non‐linear kernel (RBF – Gaussian radial basis function) implemented in a library for support vector machines – LIBSVM (Chang & Lin, 2011).

The performance of an OCSVM strongly depends on its hyperparameter selection which is used to control the learning process and in our case was a combination of a Gaussian kernel with width σ and the coefficient ν. ν represents an upper bound on the fraction of outliers (rejected target data) and is also the lower bound on the fraction of support vectors (Schölkopf et al. ²⁰⁰¹). To tune these hyperparameters we used the self‐adaptive data shifting (SDS) method proposed by Wang et al. (2018), which by shifting detected edge patterns can generate high quality pseudo target and outlier data. Parameters of edge detection used in SDS were extensively studied by Li & Maguire (2011) and in this work we used recommended values for threshold (0.1) and number of nearest neighbours (ceil(5 × ln(N)), where N is number of training samples). First, we fitted linear dimensionality reduction PCA, to measured training data (estimated parameters of Ca²⁺ events; amplitude, TTP, FDHM, FWHM, τ_decay, area under the fit) and extracted two principal components. Then we generated pseudo target and outlier data for our measured dataset of photolytic events (Fig. 1E ). The OCSVM model was trained on a dataset of measured photolytic signals and validated using generated pseudo data. The best combination of hyperparameters was found using a Bayesian optimization implemented in MATLAB (function bayesopt) with a Matthews correlation coefficient (MCC; Matthews, ¹⁹⁷⁵; Chicco & Jurman, ²⁰²⁰) as evaluation metric. The best trained classifier was then used to classify all measured events as similar (i.e. ‘no spark’) or different to the training set (i.e. ‘spark’, Fig. 1F). Very small and short Ca²⁺ events with amplitude <0.5 ΔF/F ₀ and FDHM <5 ms were also labelled as ‘no spark’.

First triggered sparks in pairs

After classification, from all first triggered Ca²⁺events by 2P photolysis of BIST‐2ETGA, the probability of triggering a Ca²⁺ spark in wild‐type (WT) and R420Q homozygous cardiomyocytes were calculated (Fig. 2A ). In control conditions we observed a higher chance to trigger a Ca²⁺spark in cells from R420Q myocytes than in WT cells (P = 0.00141). In WT cells, in the presence of 5 μmol/l cAMP there was a significantly higher chance of triggering a Ca²⁺ spark than in control condition (P = 4.8 × 10⁻⁶) or in the presence of 1 μmol/l CPA (P = 0.00034). We observed an increase of probability of triggering Ca²⁺spark in the presence of 50 μmol/l caffeine, when compared to control conditions in WT animals, but it was not significant (P = 0.0966). Triggered Ca²⁺ sparks in WT cells had significantly higher amplitudes in the presence of 5 μmol/l cAMP than when exposed to 1 μmol/l CPA (P = 1.35 × 10⁻⁹), to 50 μmol/l caffeine (P = 0.00225) or in control conditions (P = 1.21 × 10⁻⁹, Fig. 2B ). Ca²⁺ sparks in WT animal in the presence of cAMP had also significantly longer duration (P = 1.4 × 10⁻¹⁶ vs. CPA, P = 1.4 × 10⁻¹⁴ vs. control, P = 2.9 × 10⁻¹⁰ vs. 50 μmol/l caffeine, Fig. 2C ) and were wider (P = 9.8 × 10⁻¹¹ vs. CPA, P = 9.3 × 10⁻¹² vs. control, P = 3.5 × 10⁻⁶ vs. 50 μmol/l caffeine, Fig. 2D ). There was no significant difference in parameters of triggered sparks in control conditions between WT and R420Q animals.

Figure 2. Triggered first Ca²⁺ sparks in pair.

A, probability of triggering Ca²⁺ spark with 2P photolysis of BIST‐2EGTA. Data are shown as means ± 95% confidence intervals; individual points represent average per animal. B–D, estimated parameters of first triggered Ca²⁺ sparks in WT and in mutant cells. R420Q control: N = 5, n = 17, first events = 710, first Ca²⁺ sparks = 511. WT control: N = 8, n = 26, first events = 1315, first Ca²⁺ sparks = 725; cAMP: N = 8, n = 24, first events = 815, first Ca²⁺ sparks = 637; caffeine: N = 3, n = 12, first events = 666, first Ca²⁺ sparks = 458; CPA: N = 8, n = 25, first events = 1452, first Ca²⁺ sparks = 867. Horizontal black lines represent significant difference, P < 0.05.

These observations are compatible with the notion that the CICR and RyRs in R420Q myocytes have a slightly elevated Ca²⁺ sensitivity, consistent with the overall gain of function conferred by this genotype (Wang et al. ²⁰¹⁷; Yin et al. ²⁰²¹). Similarly, the activation of protein kinase A (PKA) in the presence of cAMP is expected to increase the RyR Ca²⁺ sensitivity. This may result from both the phosphorylation of the RyRs and the stimulation of the SERCA (after phosphorylation of phospholamban). In contrast, slowing down the SERCA with a low concentration of the SERCA inhibitor CPA (to minimize SR Ca²⁺ depletion) did not significantly affect the first Ca²⁺ release in a pair (but altered the second spark, presumably by slowing down SR refilling, see below).

Second triggered sparks in pairs

The second event in a pair provides information about the recovery of the local Ca²⁺ release mechanism from termination of the first event in a pair. Recovery will depend on a variety of factors, mainly on local refilling of the SR via intra‐SR Ca²⁺ diffusion and somewhat on SERCA activity, but also on recovery of the RyRs themselves from their inactivation (Szentesi et al. ²⁰⁰⁴; Sobie et al. ²⁰⁰⁵; Ramay et al. ²⁰¹¹; Brunello et al. ²⁰¹³; Poláková et al. ²⁰¹⁵; Manno et al. ²⁰¹⁷). Local recovery of CICR was tested in control conditions, in the presence of 5 μmol/l cAMP, 50 μmol/l caffeine or 1 μmol/l CPA in WT and in control conditions in R420Q cells. Only pairs of events in which a first spark was triggered were considered. We observed a trend for an increased probability of triggering a second spark in a pair in R420Q cells in control condition and also in WT cells in the presence of 5 μmol/l cAMP (Fig. 3A ). In pairs of Ca²⁺ evens, where two biological Ca²⁺ sparks were triggered, we calculated spark amplitude recovery at different intra‐pair intervals (Fig. 3B ).

Figure 3. Triggered second Ca²⁺ spark in a pair.

A, probability of triggering of a second spark in a pair at different intra‐pair intervals. B, amplitude recovery of triggered Ca²⁺ sparks at different intra‐pair intervals. Data are shown as means ± 95% confidence intervals; individual points represent average per animal. R420Q control: N = 5, n = 14–17. WT control: N = 8, n = 19–25; cAMP: N = 8, n = 19–23; caffeine: N = 3, n = 10–12; CPA: N = 8, n = 20–25. For detailed numbers of recorded Ca²⁺ events see Supporting information, Supplemental Table 1. Horizontal black lines represent significant difference, P < 0.05.

The recovery of the CICR after the first activation and inactivation is a complex phenomenon. Our data show that mechanisms making the CICR and the RyRs more Ca²⁺ sensitive (e.g. cAMP and the R420Q mutation) generally accelerate the recovery and lead to higher trigger probabilities. However, in the case of cAMP, stimulation of the SERCA and thus more rapid SR refilling may also contribute to the faster recovery. This is underscored by the observation that the SERCA inhibitor CPA slows down and reduces the recovery. The SERCA stimulation may also explain the larger amplitude of these release signals, resulting from a higher SR Ca²⁺ load. The longer spark duration in cAMP, despite SERCA stimulation, may result from spurious RyR openings during the spark decline. Differences in apparent Ca²⁺ sensitivities of CICR and the RyRs could also arise from functional or ultrastructural remodelling occurring as an adaptive process in the transgenic R420Q cardiomyocytes.

Discussion

The analysis of spontaneous Ca²⁺ sparks occurring during diastole allows examination of function of RyRs on the near molecular level in situ and in live cells. However, one has little or no control regarding the timing and location of these random and accidental events. This limitation can be partly overcome by triggering these events by 2P photolysis of caged Ca²⁺. In this work we present a novel approach for measuring the Ca²⁺ sensitivity of CICR and ryanodine receptors using a recently developed ultra‐effective Ca²⁺ caged compound (Agarwal et al. ²⁰¹⁶). In permeabilized cardiomyocytes precise spatiotemporal control over Ca²⁺ release from the SR allowed us to determine Ca²⁺ sensitivity of CICR and the RyRs as well as the trigger probability of a second spark, which is a parameter for the recovery of the RyRs after termination of the previous release.

Triggering Ca²⁺ sparks with 2P photolysis on a physiologically relevant time scale required the development of a new caged Ca²⁺ compound with a large 2P absorption cross‐section. We took advantage of knowledge gained from structure–activity relationships gained for 2P‐induced fluorescence from extended pi‐electron systems (Albota et al. ¹⁹⁹⁸; Rumi et al. ²⁰⁰⁸), and applied it to Ca²⁺ uncaging by making BIST‐2EGTA (Agarwal et al. ²⁰¹⁶). The core chromophore has a 2P cross‐section in the 350–775 GM range, significantly larger than our previous Ca²⁺ cages (range 0.01–0.6 GM (DelPrincipe et al. ¹⁹⁹⁹; Momotake et al. ²⁰⁰⁶)). Since Ca²⁺ is photoreleased in less than 1 ms, 2P irradiation can serve as a photochemical trigger with near physiological characteristics for Ca²⁺‐regulated events such as Ca²⁺ sparks (Wang et al. ²⁰⁰¹). This was not possible with previously used caged Ca²⁺ compounds, as the low 2P cross‐section meant significantly longer pulse durations (30–50 ms) were required to elicit any Ca²⁺release from the SR. Such photolytic protocols even outlasted the typical duration of the sparks themselves and thus were not suitable for triggering pairs of Ca²⁺ sparks with intrapair intervals of less than 200 ms (Lipp & Niggli, 1998; DelPrincipe et al. ¹⁹⁹⁹; Lindegger & Niggli, ²⁰⁰⁵; Ellis‐Davies, ²⁰²⁰). Further, we note the exceptional 2P cross‐section of BIST allowed us to use much lower concentrations of BIST‐2EGTA (0.05 mmol/l) compared to simpler nitro‐aromatic caged Ca²⁺ compounds such as DM‐nitrophen (up to 10 mmol/l was required).

To use of this new approach really effectively, several problems needed to be overcome. Separation between classes of purely photolytic events and true triggered Ca²⁺ sparks is challenging due to overlapping distributions of several parameters of these events (Fig. 1E and F), complicated by the fact that we have a clear identification (even though based on multiple signal parameters) only for one class, for pure photolytic events measured in the presence of 10 mmol/l caffeine and 20 mmol/l BDM (Fig. 1A ). To overcome this obstacle, we used a one‐class support vector machine (OCSVM) as a technique to precisely describe data from one class (pure photolytic signals only) and treated other signals as outliers (corresponding to Ca²⁺ sparks). The main problem with using an OCSVM is the choice of hyperparameters (σ and ν) of the classifier, which are used to control the learning process. Simple use of Ca²⁺ events with very different parameters such as photolytic signals (median of parameters of photolytic signals + 3 × median absolute deviation) was not sufficient to find a reasonable hyperplane to separate these two classes of Ca²⁺ events. To overcome this, we adopted a technique of self‐adaptive data shifting (Fig. 1E ) to create high quality and reproducible outliers (as representation of the Ca²⁺ spark class), which were then used to tune the hyperparameters of the classifier using a Mathews correlation coefficient (MCC) as metric. The MCC was preferred over the f1 score, which is a harmonic mean of the precision and recall, due to better performance in binary classification in highly imbalanced classes (Chicco & Jurman, 2020).

It was shown that each Ca²⁺ release event, either local or cell‐wide, induces refractoriness. The time course of Ca²⁺ spark recovery from refractoriness depends on a variety of parameters, such as sensitivity of ryanodine receptors to cytosolic Ca²⁺ and local sarcoplasmic reticulum refilling (Terentyev et al. ²⁰⁰²; Szentesi et al. ²⁰⁰⁴; Sobie et al. ²⁰⁰⁵; Picht et al. ²⁰¹¹; Ramay et al. ²⁰¹¹; Poláková et al. ²⁰¹⁵). In the literature, multiple mechanisms have been proposed to explain refractoriness and termination of Ca²⁺ release. The most recent ones are ‘induction decay’ (Cannell et al. ²⁰¹³; Laver et al. ²⁰¹³) or ‘pernicious attrition’ (Gillespie & Fill, 2013), which provide adequate explanations for termination of CICR. Briefly, the decline in [Ca²⁺]_SR results in a decrease of release flux during a Ca²⁺ spark, which cause the decrease of local cytoplasmic [Ca²⁺] in the dyad, resulting in a prolongation of closed time of RyRs. With RyR closed times getting longer, the probability of opening/re‐opening of RyRs within a cluster decreases and CICR becomes unlikely to continue. This process may be also modulated by many different factors from the cytosolic side, which include Ca²⁺, calmodulin and post‐translational modification of RyRs (phosphorylation, oxidation). From the SR luminal side there are findings suggesting direct Ca²⁺ modulation of RyRs (Jiang et al. ²⁰⁰⁷) or through luminal Ca²⁺ sensors (e.g. calsequestrin; Terentyev et al. ²⁰⁰²; Manno et al. ²⁰¹⁷). Impaired (shortened) refractoriness, as in the case of mutation of R420Q ryanodine receptors (Figs 2A and 3A ), might lead to arrhythmogenic diastolic Ca²⁺ releases (Lopez et al. ²⁰²⁰; Yin et al. ²⁰²¹). PKA stimulation by cAMP increased the probability of triggering of Ca²⁺ sparks (Figs 2A and 3A ), which agrees with results of spark recovery analysis using the ‘ryanodine method’ (Sobie et al. ²⁰⁰⁵) in the presence of isoproterenol (Ramay et al. ²⁰¹¹; Poláková et al. ²⁰¹⁵; Potenza et al. ²⁰¹⁹). Our analysis of Ca²⁺ spark amplitude recovery did not show any significant difference resulting from pharmacological interventions or when compared to the RyR CPVT mutation (Fig. 3B ). A reason for this might be that the real amplitude of triggered Ca²⁺ spark is partly masked by Ca²⁺ release from photolysed BIST‐2EGTA. This situation is a limitation of the 2P approach and might happen when the amplitude of triggered Ca²⁺ spark is considerably smaller than amplitude of Ca²⁺ released from caged compound, which might explain discrepancies in Ca²⁺ spark amplitude recovery observed by the ‘ryanodine method’.

It is possible that due to the complex geometry and distribution of Ca²⁺ release units in cardiac myocytes (Galice et al. ²⁰¹⁸; Jayasinghe et al. ²⁰¹⁸; Kolstad et al. ²⁰¹⁸; Sheard et al. ²⁰¹⁹; Shen et al. ²⁰¹⁹; Asghari et al. ²⁰²⁰), we might trigger Ca²⁺ releases from multiple CRUs during the 2P stimulus protocol. This is due to the fact that in our case the diffraction limited 2P volume is less than 0.2 fl (60 × objective with NA = 1.20, λ_2P = 810 nm) with ∼1 μm in axial and ∼0.5 μm in lateral dimension (Brown et al. ¹⁹⁹⁹; Zipfel et al. ²⁰⁰³; Rubart, ²⁰⁰⁴). However, a significant advantage compared to the analysis of spontaneous Ca²⁺ sparks is the fact that we are exclusively recording Ca²⁺ events from the focal plane. Furthermore, we are not limited by the duration of these experiments, as in the case of the ‘ryanodine method’. In the presence of low‐dose ryanodine, the appearance of long lasting sparks (RyRs in sub‐conductance states or multiple ryanodine molecules bound to RyR) is often observed after ∼10 min and therefore the duration of such recordings is somewhat limited (Sobie et al. ²⁰⁰⁵; Ramay et al. ²⁰¹¹). Another convenience of the photolytic approach is that we have perfect control over timing and location of the 2P pulses to trigger Ca²⁺ events. There is no need for a time‐consuming search for repetitive sparks in cells, which might be challenging especially in conditions with a low probability of observing repetitive sparks (e.g. in the presence of Ca²⁺ release blockers). These advantages are analogous to the use of 2P uncaging of glutamate at dendritic spines (Matsuzaki et al. ²⁰⁰¹), when compared to evoked release (Sabatini & Svoboda, 2000). The latter is really limited by release probability, whereas uncaging completely bypasses this restriction, allowing pure postsynaptic effects to be probed with 1 to 1 certainty (Ellis‐Davies, 2018).

In summary we introduce an experimental method which provides composite information of Ca²⁺ release probability and amplitude recovery in cardiomyocytes reflecting both SR refilling and restoration of Ca²⁺ sensitivity. More broadly, we have established that BIST‐2EGTA can be used as an extremely effective source of photoreleased Ca²⁺, as very short pulses of light (1 ms) and low powers (20 mW from a mode‐locked Ti:sapphire laser) from low concentrations of the probe (only 50 μmol/L) can induce Ca²⁺ ‐induced physiological responses that are indistinguishable from the natural events. Since Ca²⁺ is a ubiquitous second messenger in many cell types, our current work with BIST‐2EGTA indicates that new avenues of optical physiology can be examined for the first time.

Author's present address

H. Agarwal: Hitesh Agarwal, School of Pharmacy, South University, Savannah, GA, USA

Competing interests

The authors declare that they have no competing interests.

Author contributions

All of the experiments were performed at the Department of Physiology at the University of Bern. E.N., R.J., G.C.R.E.D., M.E. and A.M.G. conceived and designed the study. H.A. and G.C.R.E.D. developed the new caged compound. A.M.G. engineered the transgenic mouse model used for validation. R.J. acquired and analysed the data. R.J., E.N. and G.C.R.E.D. interpreted the data and drafted the manuscript. All authors revised the manuscript and provided intellectual feedback. All authors have read and approved the final version of this manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding statement

This project was supported by the Swiss National Science Foundation(SNSF grants 31003A 179325 and 310030 156375 to E.N., 310030 185211 to M.E.), by the NIH (GM53395 to G.C.R.E.D and 2R01HL055438‐22 to A.M.G.) and Agence Nationale de la Recherche (ANR‐19‐CE14‐0031‐01 to A.M.G.).

Supporting information

Supplemental Table 1. Numbers of triggered second Ca²⁺ events in pairs in different experimental conditions.

Statistical Summary Document

Peer Review History

Biography

Radoslav Janicek received his PhD in animal physiology from the Comenius University in Bratislava. After a postdoctoral stay in Rutgers New Jersey Medical School in Newark, he moved to Bern to the Department of Physiology of the University of Bern. His research focuses on calcium signalling in cardiac myocytes using a combination of confocal imaging, electrophysiology techniques and photolysis of caged compounds.

Data availability statement

All data used for the analysis will be made available through the public data repository Zenodo.org (DOI of dataset https://doi.org/10.5281/zenodo.4745844).