Ca²⁺ Sparks and Waves in Canine Purkinje Cells

Abstract

We have investigated the subcellular spontaneous Ca²⁺ events in canine Purkinje cells using laser scanning confocal microscopy. Three types of Ca²⁺ transient were found: (1) nonpropagating Ca²⁺ transients that originate directly under the sarcolemma and lead to (2) small Ca²⁺ wavelets in a region limited to ≈6-μm depth under the sarcolemma causing (3) large Ca²⁺ waves that travel throughout the cell (CWWs). Immunocytochemical studies revealed 3 layers of Ca²⁺ channels: (1) channels associated with type 1 IP₃ receptors (IP₃R₁) and type 3 ryanodine receptors (RyR₃) are prominent directly under the sarcolemma; (2) type 2 ryanodine receptors (RyR₂s) are present throughout the cell but virtually absent in a layer between 2 and 4 μm below the sarcolemma (Sub-SL); (3) type 3 ryanodine receptors (RyR₃) is the dominant Ca²⁺ release channel in the Sub-SL. Simulations of both nonpropagating and propagating transients show that the generators of Ca²⁺ wavelets differ from those of the CWWs with the threshold of the former being less than that of the latter. Thus, Purkinje cells contain a functional and structural Ca²⁺ system responsible for the mechanism that translates Ca²⁺ release occurring directly under the sarcolemma into rapid Ca²⁺ release in the Sub-SL, which then initiates large-amplitude long lasting Ca²⁺ releases underlying CWWs. The sequence of spontaneous diastolic Ca²⁺ transients that starts directly under the sarcolemma and leads to Ca²⁺ wavelets and CWWs is important because CWWs have been shown to cause nondriven electrical activity.

In cells devoid of t tubules such as atrial and Purkinje cells (Pcells), excitation–contraction coupling (ECC) involves Ca²⁺ release from stores located near the sarcolemma and subsequent Ca²⁺-induced Ca²⁺ release (CICR) along a lattice of sarcoplasmic reticulum (SR) enveloping the sarcomeres, which then activate myofibrils throughout the cell.^1–5 Drugs which affect SR function, such as thapsigargin and ryanodine, inhibit Ca²⁺ activation of Pcells.^2,6 Conversely, spontaneous nonpropagating Ca²⁺ release and Ca²⁺ waves cause sarcolemma depolarization in both pacemaker cells and Pcells, which can lead to nondriven electrical activity even at normal [Ca²⁺]_o.^2,7–9 Abnormal Ca²⁺ release in the network of Pcells may also be involved in lethal arrhythmias after myocardial infarction.^7,10 Previous observations suggested that micro Ca²⁺ transients, spanning only a few micrometers and traveling over short distances, initiate cell-wide Ca²⁺ waves (CWWs) which in turn induce nondriven electrical activity in a Pcell aggregate.⁷ Here, we determined the mechanistic relationships between the different subcellular spontaneous Ca²⁺ events in canine Pcells using confocal microscopy.

Materials and Methods

Eighteen aggregates of 2 to 6 cells were enzymatically dispersed from the Purkinje network of canine left ventricle (n = 9)⁷ and placed in a chamber on the stage of an inverted laser scanning confocal microscope (LSCM). Fluorescence was measured only in rod-shaped Pcells with typical junctional ends, clear striations, and membranes free of blebs.²

Measurement and Analysis of Ca²⁺ Transients

Confocal line-scans were first positioned parallel to the long axis in the cell-center (longitudinal scans) and then moved to the lateral edge or to the top/bottom edge. The transition of fluorescence between the cytosol and extracellular fluid was used to localize the sarcolemma in transverse scans and scans through cell borders.

Local variations of [Ca²⁺]_i along scan-lines were estimated from the pixel-to-pixel ratio F/Fo (F: instantaneous fluorescence; Fo: reference fluorescence) and analyzed using custom programs in IDL (IDL 5.4, Research Systems).

To mechanistically understand the processes that contribute to both propagating and nonpropagating Ca²⁺ transients, we constructed a mathematical model of release, diffusion, binding, and reuptake of Ca²⁺ in an array of (50) nodes. Ca²⁺ changes in scan line images were simulated by numerical integration of the differential equations¹¹ for all Ca²⁺ fluxes. For details see the online supplement available at http://circres.ahajournals.org.

Immunolabeling

The immunocytochemical protocol used was similar to that previously described.¹² Primary antibodies used were anti-IP₃R₁ (1:1000), an antibody that recognizes all 3 RyR isoforms, anti-RyR₂ (IgG₁; clone C3–33 and clone 34C, respectively; Affinity Bioreagents Inc; 1:500), and anti-RyR₃ and anti-RyR₁ produced and verified as described previously.^13,14 For each aggregate, serial slices (2-μm intervals) through the z-axis of the entire aggregate were imaged. Antibody label density across cells was obtained from a pixel-to-pixel average of fluorescence of 30 line arrays across the cells after correction for nonspecific fluorescence (see online supplement).

Results are expressed as mean ± SEM. Comparisons were performed on groups of data by ANOVA. The difference was significant when P<0.05 after Bonferroni adjustment.

Results

Nonpropagating Ca²⁺ Transients

Nonpropagating Ca²⁺ transients were ubiquitous throughout Pcells. Most (82%) of these events (amplitude = 1.85 ± 0.02 F/Fo; duration at half maximal amplitude [T_0.5] = 41 ± 1 ms; full width at half maximal amplitude [FWHM] = 3.2 ± 0.1 μm; rate = 0.6 ± 0.2 events per s per 100 μm; n = 524) were similar to Ca²⁺ sparks reported for rabbit Pcells.^6,15 The remaining nonpropagating Ca²⁺ transients had amplitude above 3 F/Fo (amplitude = 4.9 ± 0.2 F/Fo; T_0.5 = 43 ± 2 ms; FWHM: 2.4 ± 0.2 μm; rate = 0.13 ± 0.04 events per s per 100 μm scan; n = 115; Figure 1A). Clusters of consecutive and/or simultaneous Ca²⁺ sparks (Figure 1B) were also detected predominantly in regions below the sarcolemma and were similar to compound sparks described previously.¹⁶ Early sparks in clusters were often followed by a progressive increase in amplitude of later sparks (Figure 1B), suggesting an avalanche of multiple Ca²⁺ releases from a single or several adjacent sites with summation of Ca²⁺. However, amplitude of <5% of this compound sparks exceeded 3 F/Fo.

Figure 1.

Nonpropagating Ca²⁺ transients. A, Histograms of maximal amplitude (F/Fo_max), T_0.5 and FWHM (see text) of 639 nonpropagating Ca²⁺ transients. Binned results are expressed as percentage of total number of observations. B, Spatiotemporal distribution and time course (upper trace) of [Ca²⁺]_i during a Sub-SL “compound sparks” (see text). Insets throughout the figures show positions of the confocal plane (1) and the scan line (2) in the cell aggregate (3). Ca²⁺ concentrations are expressed as F/Fo as indicated by the pseudo-color bars. C, Examples of nonpropagating Ca²⁺ transients near the sarcolemma as detected by a 25 μm transverse line-scan across a Pcell during 4.5 s (3 consecutive frames, 1.5 s/frame); some events exhibit characteristics of single regular Ca²⁺ sparks (arrows) while others showed amplitudes up to 9 F/Fo with large spread into the cell-center. D, Typical large nonpropagating Ca²⁺ transient observed under the sarcolemma (left); note the asymmetrical spatial distribution (right trace) of the Ca²⁺ transient; image on the right shows a simulation of the transient (see text and online supplement); the curvilinear displacement of the front of the Ca²⁺ transient (see upper inset: red lines) suggests that Ca²⁺ moves by diffusion from the original site of release under the sarcolemma (see text) and disappears by active removal.

Transversal scans revealed the presence of spontaneous nonpropagating Ca²⁺ transients with various amplitudes directly under the sarcolemma (Figure 1C and supplemental Figure Is). The asymmetrical spread of these events (Figure 1D) confirmed that releases occurred directly under and against the membrane. Multiple consecutive releases could occur from the same site under the sarcolemma and produce large rises of Ca²⁺. The majority of the large (amplitude >3 F/Fo) nonpropagating events detected in our study actually occurred in this region (supplemental Figure Is).

Propagating Ca²⁺ Transients

Ca²⁺ transients also propagated as waves with linear fronts that extended from several micrometers to the full length of the scan-line (Figure 2). Two types of Ca²⁺ waves were identified: small waves or wavelets (amplitude <3 F/Fo) and large waves (amplitude ≥3 F/Fo) or CWWs.¹

Figure 2.

Propagating Ca²⁺ transients. A, Wavelets sampled by 25-μm line-scans of the lateral (a: frames 1,2,3) and top (b: frames 2,3) Sub-SL. On line-scan images, wavelets often exhibited V-shapes (eg, frames a1, b3) suggesting that Ca²⁺ events started from one site in the Sub-SL and, from there, propagated at nearly constant velocity in all directions; note the presence of Ca²⁺ sparks (yellow arrows). B, 25-μm transverse scan-line through the cell-center shows that wavelets propagate in a layer of ≈6-μm depth which may however extend occasionally deeper in the cell (frame 4); the width of the aggregate was 25 μm with 0 corresponding to the position of the sarcolemma. C, Longitudinal line-scans (25 μm) in the Pcell center reveal both typical Ca²⁺ sparks (F/Fo max = 1.9, FWHM = 1 to 2 μm) and CWWs (amplitude = 8.3 F/Fo, velocity = 147.2 μm/s); bottom figure (0.8 s by 25 μm) shows a representative example of diffusional flags (see text); note that speed is similar to that of Sub-SL wavelets (here 110 μm/s) and amplitude decreases without true front wave as shown by a-b profiles (see adjacent panel). D, Three-dimensional representation of frame 4 of Panel B illustrates large Ca²⁺ increases which are occasionally detected at the origin (white arrow) of wavelet under the membrane.

Wavelets had a short rise time, decayed exponentially after the peak, and lasted ≈150 ms (Figures 2 and 6B and Table). Frequently wavelets started under the sarcolemma after nonpropagating Ca²⁺ elevations (Figure 2B and 2D) and in areas with prominent spark activity (Figure 2A). They propagated over a limited depth from the sarcolemma into the cell: 6.2 ± 0.2 μm; n = 224; 17 cells. We denote this 6-μm layer under the sarcolemma as the Sub-SL region. The wavelets traveled in the Sub-SL region over <10 μm (Table) and never triggered large nonpropagating Ca²⁺ transients. Interestingly, we found a significant reduction of the frequency of wavelets with 2APB (3 μmol/L), a modulator of IP₃-sensitive Ca²⁺ release channels^17,18: 0.86 ± 0.26 (control, n = 31) versus 0.39 ± 0.18 event/μm/100 μm (2APB, n = 41); P<0.001. Same 2APB effect was observed previously on nonpropagating Ca²⁺ transients near the sarcolemma.¹⁹

Figure 6.

Simulations of propagating Ca²⁺ transients. Simulated CWWs (A) and Sub-SL wavelets (B) of Pcells comparing measured (a) and simulated (b) Ca²⁺ events; frame dimensions in A: 25 μm × 1.5s, in B: 25 μm × 1s. Right, Time course (c) of the measured (blue) and simulated (red) Ca²⁺ transients obtained from the corresponding line scan (see blue lines in a). Insets d, Time course of the simulated Ca²⁺ release fluxes (yellow) superimposed on the simulated Ca²⁺ transients of CWW and wavelet, respectively; the total Ca²⁺ release during the pulse was 957.5 nmol for CWW (A) and 25.7 for wavelet (B). Reproduction of the Ca²⁺ transients was obtained only if regenerative Ca²⁺ release was introduced in the model; Thr for Ca²⁺ release in the simulation was 5 nmol/L (wavelets) and 120 nmol/L (CWWs).

Characteristics of Propagating Ca²⁺ Transients in Canine Purkinje Cells.

		n	Velocity, μm/s	Amplitude, F/Fo	Duration, ms	Distance, μm	Rate, Event per s per 100 μm
Center (13 Cells)	F/Fo ≤3	91	180.5 ± 14.3	1.80 ± 0.04	73 ± 12	5.6 ± 0.4	0.61 ± 0.14
	F/Fo >3	21	60.0 ± 11.1	6.4 ± 0.8	574 ± 90	20.1 ± 1.0	0.14 ± 0.08
Sub-SL (12 Cells)	F/Fo ≤3	107	111.5 ± 9.6	1.68 ± 0.04	160 ± 19	7.7 ± 0.5	0.78 ± 0.16
	F/Fo >3	22	49.3 ± 6.8	4.8 ± 0.4	471 ± 60	16.2 ± 1.4	0.16 ± 0.06

Propagating Ca²⁺ transients detected by longitudinal scans of the cell-center and the Sub-SL were classified into events with amplitude ≤3 F/Fo and amplitude >3 F/Fo. Duration and travel distance were determined, respectively, from the time and spatial coordinates of the event in line-scan images. Rate is No. of events per second extrapolated to 100-μm scan; Velocity, averaged motion speed of event; velocity of a wave is calculated from the slope of the front on single-line-scan images; velocity of ‘wave-flag’ sampled in the cell-center (Figure 2C) is calculated by linear regression through Ca²⁺ maxima in the line-scan image. Note that diffusional flags cross the longitudinal scan-lines with angle, which leads to overestimate their velocity. Results are expressed as mean ± SEM.

CWWs typically extended from sarcolemma to sarcolemma (Figures 2C and 3), occurred at ≈5-fold lower frequency, and propagated along the aggregate at 2-fold lower velocity than wavelets (Table). CWWs lasted considerably longer than wavelets. They consistently exhibited “pseudo-plateaus” (Figures 2C, 3, and 6A) at a wide range of peak F/Fo (from 3 to 11 F/Fo; 21 waves) thus ruling out an artifact attributable to saturation of fluorescence at elevated [Ca²⁺]_i.

Figure 3.

Sub-SL wavelets trigger the CWWs. A, Longitudinal line-scan image of a wavelet (yellow arrow) preceding the initiation and propagation of a typical CWW (white arrow) in the Sub-SL B, F/Fo profiles 1, 2, and 3 of Panel A; the wavelet triggers Ca²⁺ release leading to a larger Ca²⁺ wave when [Ca²⁺]_i exceeds 1.8 F/Fo (see arrow); the large wave exhibits a plateau and a slow decay characteristic of CWWs (see text). C, F/Fo was converted into [Ca²⁺]_i using Cheng et al equation²² with dissociation constant of Fluo4 = 245 nmol/L and resting [Ca²⁺]_i = 70 nmol/L;³² in this typical example, wavelet and CWW were triggered respectively by (Thr) 5 and 80 nmol/L increase of [Ca²⁺]_i above resting [Ca²⁺]_i. D, The time course of [Ca²⁺]_i shown in profile 1 (A and B) was reproduced accurately by the sum (red trace) of a wavelet (inset: yellow trace) and a CWW (inset: blue trace); both the wavelet and CWW were simulated using the Thr values determined in C. E, Ca²⁺ transients in the gap-junction area (GJ) recorded by longitudinal scans (25 μm) through the region connecting two Pcells (C1, C2). Only raw fluorescence is shown (no ratio) so that the junctional region (GJ) is readily visible as a low fluorescence area; pseudo-colors are based on arbitrary scale with minimum and maximum equivalent to ≈1 and 4 F/Fo respectively; the panel shows the arrival of a Ca²⁺ wavelet at the border of C1 which then led to a [Ca²⁺] rise in GJ followed by a Ca²⁺ increase near the sarcolemma of C2. C2 responded by generating a large wave (Fig. 3E: F/Fo = 4.1; plateau: 180 ms; velocity: 39.8 μm/s). The latency of initiation of waves in cells of type C2 depended on the amplitude of the initial transient near the sarcolemma (here 2 to 3 F/Fo).

Transverse scans showed that small [Ca²⁺]_i increases triggered wavelets (threshold: Thr <10 nmol/L; n = 6) and wavelets triggered CWWs (Figures 3 and 4) at an ≈10-fold higher threshold (Thr≈70 to 120 nmol/L; n = 5; Figure 3C).

Figure 4.

Two-dimensional confocal Ca²⁺ images of the Sub-SL events. A and B show 2 series of consecutive 2D frames (extracted from clips in the online supplement) constructed from x/y scanning (100 scan lines, 10 μm/line, 1.2 ms/line) of cell region covering sarcolemma, Sub-SL, and a small region of the cell-center (see top inset). [Ca²⁺]_i is expressed in relative variation of F/Fo(ΔF/Fo = F-Fo/Fo). A, A wavelet propagates from the left to the right within a ≈6-μm wide region under the sarcolemma (Sub-SL). B, A similar wavelet in the same aggregate initiates (frame B4) a wave that propagates to the cell-center (see arrow in B5). The green line, superimposed to the frames, underlines the position of the sarcolemma as the boundary between intracellular fluorescence and extracellular milieu. C, Sub-SL nonpropagating Ca²⁺ transient in the same aggregate. D, Spatiotemporal Ca²⁺ profiles (indicated by corresponding frame number: B2, B3, B5, B7) of wave generation shown in B; profiles were obtained by horizontal pixel-averaging of 20 vertical lines as indicated in frame B2 (white a-b box) and the upper inset of panels A, B, C, and E: preliminary observations of wavelets in the Sub-SL region of a Pcell aggregate loaded with Fluo4 by fast video imaging (27 fps) using a high-speed spinning confocal (Nipkow) disk; 10 consecutive frames (see video clip 3 in the online supplement) show the initiation (frame 1) and propagation of a small wave in a restricted region under the SL; The vertical arrow points at the initiation site while the dotted yellow line tracks frame to frame the position of the wave front; frame size: 34 μm × 25 μm with 10 pixels/μm.

Scans through an end-to-end boundary between cells confirmed the initiation sequence of Ca²⁺ waves and that the phenomenon perpetuates from cell to cell in the aggregate (Figure 3E).

The suggestion that Ca²⁺ transients start near the sarcolemma, and from these sites, propagate through the Sub-SL and/or into the cell-center, was further corroborated by x/y scanning (supplemental Figures IVA, IVB, and IVC). The 2-dimensional confocal images showed that spontaneous nonpropagating Ca²⁺ transients (Figure 4C; see online video clips) were common in the Sub-SL and revealed the presence of small Ca²⁺ transients propagating parallel to the sarcolemma within a layer of ≈6-μm thickness (Figure 4A, 2 through 4). Amplitude and propagation of these transients were similar to those of wavelets revealed by single line-scan technique (Figure 4D). Two-dimensional confocal images showed also that wavelets were frequently accompanied by diffuse increases of [Ca²⁺]_i (“wave flag”) below the Sub-SL (see also online supplement). These “diffusional flags” were detected by single line-scans positioned further in the cell center and appeared as small events moving with same velocity than SubSL wavelets but with decreasing amplitude (see Figure 2C). On occasion, 2D animations clearly showed a wavelet initiating a CWW while traveling longitudinally in the SubSL (Figure 4B, 4 through 6; Figure 4D).

Our linescan observations were corroborated by fast confocal video imaging (Figure 4E), which indicates that Ca²⁺ wave propagation occurred on at least 2 functional subcellular levels, suggesting that elements required for propagation differ in 2 different regions.

Immunolabeling of Ryanodine and IP₃ Receptors

We determined the nature and distribution of SR Ca²⁺ release channels in these Pcells. First, we found no labeling for IP₃R₂ and RyR₁ (results not shown). However, colabeling with IP₃R₁ and RyR₂ antibodies was positive and revealed 2 distinct regions: a layer of ≈2 μm thick with IP₃R₁ label (Figure 5A.b and 5B; green) existed directly under the sarcolemma while intense RyR₂ labeling dominated most of the cell including the lateral IP₃R₁-positive layer (Figure 5A.b and 5C; red). Analysis of IP₃R₁- and RyR₂-positive regions showed that only 5% of the corresponding pixels overlapped. Furthermore, Figure 5A.b and 5C illustrate a novel and typical feature of antibody staining of Pcells, in that label for both RyR₂ and IP₃R₁ was extremely sparse in a continuous layer of 2 μm thick below the IP₃R₁-positive layer (see arrows). This apparent void of Ca²⁺ channels was observed with RyR₂ antibody but not with a RyR antibody that recognized all isoforms of the channel (Figure 5A). We found that this void was filled by a specific RyR₃ antibody which actually labeled a layer of 7-μm thickness below the sarcolemma (Figure 5C) and was absent in the cell-center (Figure 5A.c, 5B.c, and 5C).

Figure 5.

Localization of SR Ca²⁺ channels (Ryanodine and IP₃ receptors) in canine Pcell. A, Optical sections of Pcells labeled respectively for all RyR isoforms (A.a), IP₃R₁ and RyR₂ (A.b), and RyR₃ (A.c). Fluorescence signals were corrected for nonspecific fluorescence. a shows that a RyR antibody, which probes all RyR isoforms, stains the cell uniformly and exhibits a transverse striation of ≈2 μm. Costaining shown in b reveals that IP₃R₁ (green) forms a thin layer (1.91 ± 0.49 μm) directly under the sarcolemma while RyR₂ (red) is uniformly distributed throughout the cell and exhibits a striation with 1.9- to 2-μm periodicity. Neither IP₃R₁ nor RyR₂ were detected in a region of 2-μm width (1.9 ± 0.2 μm) between the sarcolemma and the cell-center (see white arrow). In c, Pcell stained for RyR₃ (blue) exhibits a similar 1.9- to 2-μm transversal striation but exclusively located at the edge of the cell, in a region extending ≈6 to 7 μm from the sarcolemma; note the absence of void as observed in b. B, fluorescence intensity was averaged for IP₃R₁, RyR₂, RyR₃ antibodies along transverse profiles through the cells; each profile was obtained from pixel-to-pixel averaged arrays of 30 lines. C, Distribution of IP₃R₁, RyR₂, and RyR₃ in the Sub-SL region; half of each profile of B was corrected such that 0 μm distance corresponds to the fluorescence boundary of the cell; this procedure shows clearly that RyR₂ is distributed almost uniformly throughout the cell whereas IP₃R₁ is concentrated laterally in a thin layer under the SL; the normalized spatial distributions of RyR₂ and IP₃R₁ confirms the presence of a void (see arrow) which is filled by RyR₃ (distributed from the sarcolemma to 7 ± 0.8 μm in the cell; blue). Note that RyR₃ and RyR₂ overlap below the void.

Properties of the Generators of Ca²⁺ Transients

Using model simulations, we studied whether the large Ca²⁺ waves are indeed generated by nonpropagating Ca²⁺ releases near the sarcolemma, such as has been shown in rabbit Pcells, or are in fact caused by propagating Ca²⁺ releases. We also used the simulations to identify the factors that determine the properties of wavelets and CWWs.

Sparks near the sarcolemma, in the Sub-SL and cell-center could be simulated accurately using similar parameters of the release (release time ≈40 ms) and uptake functions (see online supplement for details). The correlations between simulated and experimental data were robust for all events (R²>0.97; Figure 1D and supplemental Figure IIs). The Ca²⁺ diffusion coefficient (D_eff) required to fit such sparks was 12.1 ± 0.3 μm²/s in both the Sub-SL and cell-center; we used this value to simulate Ca²⁺ wavelets and CWWs (Figure 6). Large Ca²⁺ sparks were reproduced accurately by assuming a larger Ca²⁺-release flux and a longer Ca²⁺-release time (up to 100 ms), without a change in other parameters.

Large Ca²⁺ waves are expected to travel faster than small waves; actually, we found the opposite when we compared CWWs with wavelets: wavelets propagated on average twice as fast as CWWs (Table). The simulations ruled out that this observation was caused by differences in D_eff (see above) or extrusion kinetics (online supplement). The spacing^20,21 of Ca²⁺ channels (Figure 5A.a) appeared to be similar in the Sub-SL and cell-center and cannot, therefore, explain our observation either. Finally, we tested whether the observed difference in velocity between wavelets and CWWs was caused by different Ca²⁺ thresholds of the generators. The velocity appeared to be inversely proportional to Thr (see supplemental Figure IIIs), and CWW propagation required a Ca²⁺ threshold up to 20-fold higher than the one for wavelet propagation consistent with the experimentally determined values (Figure 3). In addition, the simulations suggested 2 other aspects of the unique and distinct nature of CWWs as compared with wavelets (Figure 6): a 10-fold longer lasting Ca²⁺ release occurred during a CWW (200 to 500 ms) compared with that of either wavelets or sparks (20 to 45 ms), whereas the calculated total Ca²⁺ release in CWWs was 40-fold larger (Cf. insets Figure 6).

Discussion

Ca²⁺ Sparks

We show here that, at normal [Ca²⁺]_o, Ca²⁺ sparks occur ubiquitously throughout canine Pcells with characteristics similar to those of ventricular myocytes,^22,23 including the presence of a subpopulation of large sparks. Large sparks in this study appear, however, wider and larger than those observed in rat cardiac myocytes.^24–26 The presence of compound sparks¹⁶ with, on average, amplitude <3 and large width in the population of nonpropagating Ca²⁺ transients may explain why, in Pcells, large nonpropagating events were slightly narrower than small events (see FWHM above).

Accuracy of the correlation between amplitude and localization of Ca²⁺ sparks is limited because local Ca²⁺ releases can occur outside the confocal plane. Nevertheless, although large single sparks were seen occasionally in the cell-center, the majority was found directly under the sarcolemma and in the SubSL region. Our observations differ from those in rabbit Pcells, where sparks are exclusively found directly under the sarcolemma.^6,15 The large sparks of canine Pcells are similar to the large 2APB-sensitive Ca²⁺ events reported in rabbit portal vein myocytes.¹⁸ The compound sparks in the Sub-SL and repetitive Ca²⁺ spark generation from single sites directly under the SL (see online supplement) suggest that both near-synchronous activation of multiple Ca²⁺ release units²⁰ and rapidly repetitive activation in one site may occur in these regions.

This is consistent with the hypothesis that activation of IP₃R may recruit adjacent Ca²⁺ channels including IP₃Rs and RyRs (or vice versa);^18,27 our findings that these 2 types of SR-Ca²⁺ channels coexist (Figure 5) and that Ca²⁺ events are sensitive to 2APB^17,19 make it probable that this IP₃R/RyR interaction indeed occurs in canine Pcells.

Ca²⁺ Transients in the Cell-Center: Diffusion or Propagation?

Cordeiro et al observed no spontaneous Ca²⁺ release centrally in rabbit Pcells although RyRs were present,^6,15 and thus concluded that central RyRs were “silent.” Our observations do not support the same conclusion, as we found that spontaneous Ca²⁺ sparks were common in the cell-center (eg, Figure 2), thus showing that RyRs are active in canine Pcells. This species difference in spontaneous Ca²⁺ release may also explain why canine Pcell aggregates and fibers exhibit nondriven electrical activity^2,7 whereas rabbit Purkinje fibers do not.²⁸

Cordeiro et al proposed that the large central Ca²⁺ elevation evoked by the action potential (AP) in rabbit Pcells resulted only from diffusion of Ca²⁺ released from peripheral sarcolemmal sites.^6,15 Our simulations of Ca²⁺ release directly under the sarcolemma (Figure 1C) are consistent with this conclusion with respect to large sparks.^6,15 In contrast, the spatiotemporal Ca²⁺ distribution of both wavelets and CWWs in canine Pcells (Figure 6) could be reproduced only by incorporating the propagation of Ca²⁺ release from node to node in the model, confirming that propagated CICR is responsible for these Ca²⁺ waves. Furthermore, our simulations show that the lower velocity of CWWs requires that Thr of Ca²⁺ release elements involved in the propagation be an order of magnitude higher than that of wavelets; this was consistent with our experimentally determined values (Figure 3). Finally, the typical plateau of CWW as well as the simulated Ca²⁺ release function are novel findings for a cardiac cell and suggest a distinct Ca²⁺ release mechanism. During CWWs, one or both of the following may occur: the release channels open completely, but the flux declines because of a decreased gradient across the channel.²⁹ Alternatively, irreversible channel opening could be induced by an interaction between the permeant ion and the channel, similar to the mechanism that has been proposed for skeletal muscle.³⁰ The determination of such mechanism is beyond the scope of this study. However, once we understand the mechanism of this persistent Ca²⁺ release, we should be able to reduce it in the intact Pcell. Such a reduction in the amplitude of CWWs and thus the depolarization that accompany them would be antiarrhythmic.

Mechanism of Initiation of Wavelets and CWWs

Spontaneously occurring Wavelets start commonly after sparks directly under the sarcolemma or in the Sub-SL (Figure 2), and, in turn, initiate CWWs (Figures 3 and 4). The effect of 2APB observed in this study, and in the previous study of Boyden et al,¹⁹ suggests that IP₃R-mediated Ca²⁺ release is instrumental in initiating/modulating Ca²⁺ wavelets. We never observed wavelets, which arrived at the sarcolemma and subsequently induced Ca²⁺ release directly under the sarcolemma. This directional asymmetry suggests that elements propagating wavelets and those causing Ca²⁺ sparks directly under the sarcolemma are functionally distinct. Furthermore, for adequate simulation of sparks near the sarcolemma, a 2-fold longer Ca²⁺ release pulse (supplemental Figure IIs) was required compared with that needed for wavelets (Figure 6), suggesting that Ca²⁺ transients directly under the sarcolemma and in the Sub-SL result from different Ca²⁺ release mechanisms.

Large CWWs span the entire cell suggesting that their underlying SR-Ca²⁺ release elements are ubiquitous. The amplitude duration and speed of the CWWs requires that the SR-Ca²⁺ release elements that mediate CWWs have a high CICR threshold and release a large Ca²⁺ flux for hundreds of ms. Faster and smaller low-threshold wavelets were found in the Sub-SL, suggesting that the SR-Ca²⁺ release elements that reside in this region have a low CICR threshold and release a small Ca²⁺ flux for tens of ms. These 2 functionally distinct Ca²⁺ release elements must overlap in the Sub-SL, where both CWWs and wavelets can exist. This arrangement predicts that the probability for wavelets to trigger CWWs is small, as reported previously² (Table). On the other hand, when the high CICR threshold of the CWW generators has been surpassed, Ca²⁺ release elements in the SR network will propagate the wave as far as this network reaches.

In summary, we show here that canine Pcells contain 3 functionally distinct Ca²⁺ release systems: system (1) is restricted to a thin layer (2 μm) directly under the sarcolemma and apposed to system (2) in the Sub-SL; system (2) partially overlaps with system (3) that drives Ca²⁺ release from sarcolemma to sarcolemma.

Ca²⁺ Release Channel Elements

We demonstrate here that a sophisticated triple-layered system of SR-Ca²⁺ release channel architecture underlies the above-proposed hierarchy of Ca²⁺ wave generation. Like in atrial cells,^3,4 we observed IP₃R, RyR₂ under the sarcolemma. However, different from atrial cells, the IP₃R₁ isoform was detected in Pcells. RyR₂ formed a clear striated pattern in the cell, similar to the pattern shown in rabbit Pcells,^6,15 as well as directly under the membrane near IP₃R₁s, but were virtually absent in a 2-μm layer (“void”) below the sarcolemma, in the subSL region. This void was specific for RyR₂ because an antibody that recognized all three RyR isoforms showed no void. In fact, RyR₃ labeling was found in high density filling the void between RyR₂ and IP₃R₁ (Figure 5).

Functional Implications

Although a detailed pharmacological analysis of Ca²⁺ transient generators in canine Pcells is beyond the scope of this study, our findings reveal that Ca²⁺ activation in spontaneous Pcells differs substantially from that of ventricular myocytes. It can be explained in the following way (Figure 7):

Figure 7.

Structural model for Ca²⁺ activity in Canine Pcells. Based on a model of RyR₃, RyR₂ and IP₃R₁ distributions (see upper diagram) and the functional model of Ca²⁺, we propose a triple layered system of SR Ca²⁺ channels underlying activation of canine Pcells. A first layer of IP₃R₁ (green squares) directly under the membrane overlaps the outer border of a second layer of RyR₃ (Sub-SL; see blue circles). The cell-center contains very little IP₃R₁ and virtually no RyR₃. The third array of Ca²⁺ release channels comprises RyR₂ (dark red circles). Some RyR₂ are present in the IP₃R₁ layer; RyR₂ density is high throughout the cell-center, but RyR₂ as well as IP₃R₁ are absent in the void (V). Below the void, RyR₂ and RyR₃ overlap. Diagrams represented on the upper part indicate the relative distribution of each Ca²⁺ channel; density scales are arbitrary and numbers are approximate sizes in μm; SL: sarcolemma. Working hypothesis for Ca²⁺ activation. (1) The IP₃R₁-RyR₃ combination generates local Ca²⁺ release directly under the membrane (event 1 in the panel), which triggers (2) Ca²⁺ release by other RyR₃s if the low-threshold of RyR₃ channels is exceeded (center of green circles); the ensuing Ca²⁺ release by RyR₃ channels propagates through the RyR₃ mesh in the Sub-SL as Ca²⁺ wavelets (green circles; event 2); (3) if the Ca²⁺ transient of wavelets exceeds the higher threshold of RyR₂ channels, RyR₂-mediated Ca²⁺ release is initiated (center of red circles). The RyR₂ channels release a large amount of Ca²⁺ and trigger both RyR₂ and RyR₃ as well as IP₃R₁ mediated Ca²⁺ release, thereby causing large CWWs (red circles; event 3).

First, a layer (2 μm) directly under the sarcolemma contains both IP₃Rs and RyR_2,3, which are either separately or in combination responsible^18,27 for the spontaneous large sparks directly under the sarcolemma. Second, the Sub-SL (6 μm thick from the sarcolemma) is a layer of RyR₃s, which partially overlaps the previous IP₃R/RyRs layer; RyR₃s will generate sparks in the unstimulated Pcell. Because of their low threshold, RyR₃s will readily respond to Ca²⁺ release directly under the sarcolemma by the generation of Sub-SL wavelets. RyR₃s indeed show spontaneous Ca²⁺ activity at normal diastolic [Ca²⁺]_i when expressed in HEK cells,³¹ and pure RyR₃ release is more sensitive to caffeine than other RyR isoforms.³¹ Their distribution would make them a source of sparks and wavelets in the Sub-SL if they also exhibit a low threshold for CICR in canine Pcells.

Third, RyR₃s overlap with widely distributed RyR₂s, thus allowing RyR₂s to respond to Ca²⁺ release from the RyR₃ Sub-SL network and generate CWWs. RyR₂ channels release a large amount of Ca²⁺ thereby recruiting all available Ca²⁺ release channels in the generation of CWWs. This model could serve as a safe unidirectional system that ensures organized activation of Pcells in response to an action potential.

In conclusion this study provides novel functional and structural evidence for a triple layered system of Ca²⁺ activation in canine Pcells involving IP₃Rs, RyR₃s, and RyR₂s explaining a well recognized feature of the Pcell aggregate: the nondriven electrical activity.⁷