Ultrafast multicellular calcium imaging of calcium spikes in mouse beta cells in tissue slices

Abstract

Background

The crucial steps in beta cell stimulus‐secretion coupling upon stimulation with glucose are oscillatory changes in metabolism, membrane potential, intracellular calcium concentration, and exocytosis. The changes in membrane potential consist of bursts of spikes, with silent phases between them being dominated by membrane repolarization and absence of spikes. Assessing intra‐ and intercellular coupling at the multicellular level is possible with ever‐increasing detail, but our current ability to simultaneously resolve spikes from many beta cells remains limited to double‐impalement electrophysiological recordings.

Methods

Since multicellular calcium imaging of spikes would enable a better understanding of coupling between changes in membrane potential and calcium concentration in beta cell collectives, we set out to design an appropriate methodological approach.

Results

Combining the acute tissue slice method with ultrafast calcium imaging, we were able to resolve and quantify individual spikes within bursts at a temporal resolution of >150 Hz over prolonged periods, as well as describe their glucose‐dependent properties. In addition, by simultaneous patch‐clamp recordings we were able to show that calcium spikes closely follow membrane potential changes. Both bursts and spikes coordinate across islets in the form of intercellular waves, with bursts typically displaying global and spikes more local patterns.

Conclusions

This method and the associated findings provide additional insight into the complex signaling within beta cell networks. Once extended to tissue from diabetic animals and human donors, this approach could help us better understand the mechanistic basis of diabetes and find new molecular targets.

1. INTRODUCTION

Pancreatic beta cells and their function have been extensively studied for decades, mostly for their decisive role in deranged glucose homeostasis in diabetic states. ¹ , ² , ³ First reports of the electrical activity of beta cells demonstrated that beta cells are electrically excitable, ⁴ , ⁵ , ⁶ and that their electrical activity, evoked by insulin secretagogues, exhibits a distinct and complex pattern. ⁷ , ⁸ , ⁹ Starting from the maximal repolarization potential (V _MR), a slow ramp depolarization by ~5 mV triggers a rapid depolarization to a plateau potential V _P (~10 mV more positive than the V _MR), from which a burst of rapid (50–100 ms) 10–20 mV depolarizations occur. These rapid fluctuations are commonly termed spike potentials (often simply abbreviated as spikes), but also other terms are used, such as calcium spikes, since these spikes are dependent on extracellular Ca²⁺ ions, and action potentials, since their duration and shape resemble action potentials in neurons. The initial frequency of the spike potentials during a burst is ~10 Hz, and the frequency of the spikes gradually decreases during the slow repolarization of the V _P, reaching finally the V _MR and a spike‐free period. During this silent period, the slow ramp depolarization back to threshold potential triggers the next burst, resulting in a regular bursting pattern. Studies reported different values for the V _MR, threshold potential, V _P, and spike amplitude, most probably due to different methodological approaches. ⁸ , ⁹ , ¹⁰ There are ample data demonstrating that in beta cells, the glucose concentration is effectively transduced into membrane potential (MP) activity. More specifically, increasing glucose concentrations increases the fraction of time at V _P relative to the silent period (commonly termed also active time or duty cycle), and this has been reported in both microdissected and isolated mouse islets, ⁸ , ¹⁰ , ¹¹ , ¹² human isolated islets, ¹³ as well for mouse islets in vivo. ¹⁰ , ¹⁴ Surprisingly, only a few studies reported the effects of glucose on spike potentials. While increasing the stimulatory glucose seems to increase the frequency of the spikes, ⁹ the amplitude and the shape of spikes seem to remain unaffected. ⁸

Already at about the time of the first descriptions of MP dynamics in beta cells, the decisive role of Ca²⁺ in insulin release was noted. ¹⁵ Over the following four decades, use of calcium dyes has generated a detailed description of glucose‐induced collective dynamics. ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ More specifically, glucose stimulation triggers an initial increase in intracellular calcium concentration ([Ca²⁺]_i) that is followed by characteristic oscillations in [Ca²⁺]_i that are superimposed on a plateau [Ca²⁺]_i level. These so‐called fast oscillations have been shown to directly correspond with bursts of spikes ¹¹ , ¹⁷ , ²² , ²³ and insulin secretion pulses. ²⁴ , ²⁵ Analogously to the MP changes, glucose concentrations effectively transduce into a change in [Ca²⁺]_i activity, increasing the total active time of fast oscillations as the glucose concentration increases. ¹⁸ , ¹⁹ , ²⁶ Moreover, the [Ca²⁺]_i oscillations are coordinated between beta cells, forming waves of [Ca²⁺]_i that follow depolarization waves and spread over islets in a repetitive manner. ¹⁶ , ¹⁸ , ²⁷ , ²⁸ The propagation of these waves is facilitated by gap junctions between beta cells composed of connexin 36 (C×36). ²⁹ , ³⁰ They act as the key synchronizing agent that enables harmonious functionality across heterogeneous beta cell populations and coherent pulsatile insulin secretion patterns. ² , ³¹

Strikingly, a [Ca²⁺]_i correlate of the MP spikes is missing in the abovementioned studies. Although simultaneous measurements of [Ca²⁺]_i and MP demonstrated that the [Ca²⁺]_i oscillations closely follow the V _P, ²² , ²³ , ³² , ³³ only a few studies reported [Ca²⁺]_i spikes that resemble the MP spikes during bursting. ³⁴ , ³⁵ , ³⁶ , ³⁷ While this pioneering work demonstrated that INS‐1 cells, cultured beta cells, and beta cells in isolated islets are capable of very fast activity in response to glucose stimulation, a systematic analysis of [Ca²⁺]_i spikes from multiple bursts in coupled beta cells in islets is still lacking and the effect of increasing glucose concentration on [Ca²⁺]_i spikes, as well as the extent of coupling of [Ca²⁺]_i spikes between different cells remain to be assessed.

To address these questions, we resorted to a [Ca²⁺]_i reporter dye with high quantum yield, that, in combination with state‐of‐the‐art confocal imaging setup, allowed unprecedented temporal and spatial resolution sufficient to resolve [Ca²⁺]_i spikes in a confocal setting. Moreover, we provide a direct experimental link between the MP and [Ca²⁺]_i dynamics by employing simultaneous patch‐clamp recording and imaging, describe how [Ca²⁺]_i spikes are interlinked with [Ca²⁺]_i oscillations (i.e., bursts), quantify the degree of coupling between [Ca²⁺]_i spikes in different cells, assess the effect of glucose concentration on the behavior of [Ca²⁺]_i spikes in many cells simultaneously, and characterize the effect of two modulatory agents on [Ca²⁺]_i dynamics.

2. RESULTS

2.1. Initial glucose‐induced beta cell recruitment evolves into coordinated [Ca²⁺]_i bursting

We used a high sampling rate (41 Hz) line scan to demonstrate the evolution of collective beta cell [Ca²⁺]_i activity following glucose stimulation (Figure 1). In an individual beta cell, the [Ca²⁺]_i evolved from an initial tonic [Ca²⁺]_i increase to an oscillatory [Ca²⁺]_i pattern with bursts of spikes. We note that in some cells the initial [Ca²⁺]_i transient lacked clear [Ca²⁺]_i spikes (Figure 1B, asterisk). Characteristically for the 1^st phase of glucose response, beta cells initiated at different times, resulting in gradual recruitment of cells within an islet. We binarized both the bursts and the [Ca²⁺]_i spikes and used this to (i) display combined cellular activity as raster plots (Figure 1D) and (ii) to calculate the average coactivity between cell pairs as a measure of intercellular coordination (Figure 1F). A coactivity value of 1 indicates perfect overlap, whereas a value of 0 indicates that no cell pair was simultaneously active within a given time interval. We calculated the coactivity separately for bursts and [Ca²⁺]_i spikes (labeled in Figure 1 as black and red, respectively). The coactivity of bursts gradually increased due to glucose‐induced activation of cells that, once activated, displayed high coordination. For [Ca²⁺]_i spikes, on the other hand, coactivity only occasionally peaked, indicating that only some spikes aligned between active cells. A qualitatively similar pattern of [Ca²⁺]_i spiking was visible in the plateau phase, characterized by global and coordinated bursting (Figure 1E–G, black) during which high spike coactivity was detected soon after the onset of a burst and declined later during a burst (Figure 1E–G, red). Due to technical limitations of [Ca²⁺]_i imaging at a high sampling rate, we limited the qualitative and quantitative description of spike characteristics to shorter intervals in the plateau phase in the following parts of the paper.

FIGURE 1.

1^st and 2^nd phase of beta cell activity following glucose stimulation sampled at high frequency. (A) [Ca²⁺]_i dynamics from 5 beta cells during stimulation with 12 mM glucose. Beta cells are recruited during the 1^st phase, whereas the plateau phase is characterized by a coordinated bursting activity. Gray areas indicate sections in panels (B, C). (B, C) Zoom‐in on cells from panel (A), revealing [Ca²⁺]_i spikes during both the 1^st and the plateau phase. Asterisk indicates an oscillation in [Ca²⁺]_i without spikes. (D, E) Raster plots indicating binarized activity of spikes (red) and bursts (black) for the cells in panels (B, C). (F, G) Coactivity of spiking (red) and bursting (black) activity for the cells in panels (B, C).

2.2. [Ca²⁺]_i spikes follow spikes in membrane potential

To correlate the [Ca²⁺]_i signal with the electrical changes, we monitored membrane potential (MP) activity from a beta cell using whole‐cell patch clamp while simultaneously measuring [Ca²⁺]_i dynamics from its neighbors (Figure 2). A faster sampling (>150 Hz) allowed for a reliable [Ca²⁺]_i description. The [Ca²⁺]_i signal strongly resembled the MP burst (Figure 2A). It consisted of an increase to a [Ca²⁺]_i plateau with superimposed spikes, followed by a gradual return of the signal to the initial level during which the spiking activity gradually ceased. Due to the strong correlation with the MP signal, we term the superimposed [Ca²⁺]_i changes as “[Ca²⁺]_i spikes,” and the whole fast [Ca²⁺]_i oscillation with superimposed spikes as a “burst of [Ca²⁺]_i spikes,” to be consistent with the broadly accepted terminology describing MP dynamics. Consistent with previous reports, we observed coordinated and space‐dependent time lags between the onsets of bursts (Figure 2A3, delays indicated by asterisks), consistent with waves of [Ca²⁺]_i spreading over the islet. ¹⁸ , ¹⁹ The data were binarized and individual MP and [Ca²⁺]_i, spikes were assigned a relative time along the course of a burst (Figure 2B). We observed that the spiking activity changed along the course of a burst (Figure 2C–E). At the onset of a burst, the MP spike frequency was on average 10.5 ± 0.7 Hz, and it gradually decreased reaching only about a half of the initial value (4.3 ± 0.8 Hz) 1 s after burst onset. The frequency of [Ca²⁺]_i spikes practically completely overlapped with the MP signal (Figure 2C). The duration of spikes differed between the two signals: For MP, the average duration was 45 ± 4 ms initially and gradually prolonged to 55 ± 12 ms after 1 s. The [Ca²⁺]_i spikes, on the other hand, were initially on average 60 ± 3 ms long and increased in duration to about twice the initial value (102 ± 5 ms) with a time course comparable to the MP signal. Finally, we calculated the active time of spikes (i.e., their duty cycle), corresponding to percentage of time populated by the spikes and measured at their half‐amplitude. ¹⁶ , ¹⁸ , ²² The active time of MP spikes was initially 0.31 ± 0.02, and it shortened to about two thirds of the initial value 1 second after burst initiation (0.19 ± 0.05), while for [Ca²⁺]_i spikes the active time was 0.55 ± 0.02 at the burst initiation and decreased to 0.36 ± 0.02 1 s after burst initiation.

FIGURE 2.

Simultaneous measurement of MP and [Ca²⁺]_i dynamics from beta cells. (A) Islet of Langerhans, loaded with the calcium reporter dye Calbryte 520 AM @ 512 × 512 pixels (A2) and simultaneously recorded transmission light image (A1) showing the position of the recording pipette. White broken line indicates the position of the line scan for the [Ca²⁺]_i imaging, and yellow squares indicate selected ROIs indicative of beta cells. Numbers correspond to cells in panel (A3), demonstrating MP from a cell near the line scan (red trace) and the simultaneously measured [Ca²⁺]_i activity in 8 different beta cells (black traces) during stimulation with 12 mM glucose. Note spatially coordinated time delays of burst onsets, indicated with asterisks. (B) Individual [Ca²⁺]_i spikes were detected at half‐maximal amplitude, as indicated with red lines. (C–E) Changes in spike frequency (C), duration (D) and active time (E) during the time course a burst, for MP (red) and [Ca²⁺]_i (black). Pooled data from two experiments involving 16 cells for [Ca²⁺]_i imaging and two individually patch‐clamped cells, during 18 bursts (N = 165 spikes for MP, and N = 1166 spikes for [Ca²⁺]_i). Shown are means (colored line) and SEM (colored area) for data divided into 10 bins.

2.3. The effect of glucose concentration on [Ca²⁺]_i spikes

To study the effect of glucose concentration on [Ca²⁺]_i spikes, we stimulated beta cells with either a physiological (8 mM) or a supraphysiological (12 mM) glucose concentration while recording the [Ca²⁺]_i activity on a larger set of experiments (Figure 3). The choice of concentrations was based on previous observations demonstrating that in vivo fasting glucose in NMRI is mice 5–6 mM, the non‐fasting values range 8–12 mM glucose ²⁶ , ³⁸ and peak close to 12 mM post‐ipGTT. ³⁹ , ⁴⁰ Individual [Ca²⁺]_i spikes and burst were extracted and characterized (Figure 3B–G). At the beginning of the burst, the spike frequency was lower by about 1.5 Hz in higher glucose (9 ± 0.3 Hz in 8 mM and 7.4 ± 0.2 Hz in 12 mM) while the duration was longer by about 20 ms (46 ± 2 ms in 8 mM and 65 ± 2 ms in 12 mM glucose). During the time course of a burst, we observed the same trend irrespective of the concentration: the frequency decreased by about 2.5 Hz (6.6 ± 0.4 Hz in 8 mM and 5.1 ± 0.2 Hz in 12 mM 1.8 s after burst onset), while the duration increased by about 20 ms (67 ± 3 ms in 8 mM and 86 ± 2 ms in 12 mM 1.8 s after burst onset). The active time of spikes at the burst onset was slightly higher in higher glucose (0.36 ± 0.01 in 8 mM and 0.40 ± 0.01 in 12 mM glucose). The curves describing the evolution of active time dissociated soon after the burst initiation, remaining at about the same active time in 12 mM glucose (0.39 ± 0.01 2.5 s after burst onset) while decreasing by about 0.20 in 8 mM (0.29 ± 0.02 2.5 s after burst initiation). In analogy, we measured the same parameters also for the bursts. The median value of burst duration was unaffected by glucose concentration (1^st quartile/median/3^rd quartile (Q1/M/Q3): 1.2/1.9/2.8 s for 8 mM and 1.1/1.5/2.8 s for 12 mM glucose), whereas an increased frequency (Q1/M/Q3 0.03/0.05/0.07 Hz for 8 mM and 0.07/0.33/0.60 Hz for 12 mM glucose) effectively increased burst active time by >3 fold (Q1/M/Q3 0.05/0.11/0.17 for 8 mM and 0.31/0.40/0.50 for 12 mM glucose).

FIGURE 3.

[Ca²⁺]_i bursts and spikes: the effect of glucose concentration. (A) An islet of Langerhans, loaded with Calbryte 520 AM, during a high temporal and spatial resolution line scan (A1). Broken line indicates the position of the line scan, and yellow squares indicate ROIs that correspond to individual beta cells. Scale bar 50 μm. Numbers correspond to beta cells in panel (A2), showing repetitive bursts of fast oscillations detected by high‐resolution [Ca²⁺]_i imaging. (B–G) Spike (B) and burst (E) frequency, spike (C) and burst (F) duration, and spike (D) and burst (G) active time, for stimulation with 8 mM (red) and 12 mM (blue) glucose. Pooled data from 3066 spikes and 248 bursts from 72 beta cells from 10 islets from 3 mice. Shown are means and SEM for spikes and quartiles for bursts, data for spikes were divided into 10 bins. ***p < 0.001, *p < 0.05 (Mann–Whitney test).

Next, we assessed the effect of glucose concentration on the shape of the [Ca²⁺]_i spikes. To this aim, we superimposed [Ca²⁺]_i spikes relative to their temporal position along the burst, separately for the two stimulatory concentrations (Figure 4). The shape of the spike consisted of a faster upstroke and a slower decay of the signal, and this general feature was unaffected by either the glucose concentration or by the position along the burst (Figure 4B,C). Detailed analysis revealed that the upstroke of the signal became slower with burst duration (Figure 4B,C), resulting in a progressive delay of the [Ca²⁺]_i peak and a prolongation of the [Ca²⁺]_i spikes, corroborating previous analysis (compare Figure 3C). Of note, due to binarization that is based on the half‐amplitude, the superimposed spikes in Figure 4 are aligned at time points corresponding to half‐amplitude. In this scenario, the slower upstroke was seen in the Figure 4 as a left shift of the beginning of [Ca²⁺]_i upstroke and a right shift of the [Ca²⁺]_i peak.

FIGURE 4.

The effect of the glucose concentration on the shape of [Ca²⁺]_i spikes. (A) [Ca²⁺]_i activity from a beta cell during stimulation with 12 mM glucose. Colors code time intervals along the course of a burst, and the same color code is used in panels (B, C). (B, C) Superimposed [Ca²⁺]_i spikes during stimulation with 8 mM (B) and 12 mM glucose (C). Colors code time intervals after burst onset (binned data into 1.5‐s intervals). Pooled data from 3069 spikes from 69 beta cells from 10 islets from 3 mice. Shown are means (colored line) and SEM (colored area) for data divided into 40 bins.

2.4. Spatiotemporal coordination of [Ca²⁺]_i spikes

Next, we investigated the spatiotemporal coordination of spikes in different beta cells, that is, if cells exhibit a coordinated [Ca²⁺]_i spiking activity (Figure 5). Focusing on a particular burst of [Ca²⁺]_i spikes, the nearby cells displayed a similar [Ca²⁺]_i spiking pattern, whereas [Ca²⁺]_i spikes in more distant cells were inherently limited by the delays in bursts, such that only during the overlapping parts of bursts the [Ca²⁺]_i spikes occurred in synchrony (Figure 5A middle panel). We used coactivity as a measure of coordinated activity. ⁴¹ In Figure 5, high coactivity spanned over the 12 cells that were up to 120 μm apart for periods of overlapping bursting activity and was limited to a few cells (30–60 μm distance) elsewhere, as set by delays and extent of overlap of the bursts (Figure 5A, upper panel). In other words, at a time point when first cells were activated, high coactivity was limited to nearby cells (left matrix in Figure 5A, high coactivity near the diagonal of the matrix). When additional cells were recruited (and the former cells were still in the bursting phase), the matrix changed to high coactivity even at larger distances (middle matrix in Figure 5A), indicating that coordinated spiking was present in many cells simultaneously. Conversely, later in the stage when the first cells already ceased their activity and the rest were still active (right matrix in Figure 5A), the coactivity was again limited to the nearby cells.

FIGURE 5.

Comparison of intercellular activity in 8 mM and 12 mM glucose for [Ca²⁺]_i bursts and spikes. (A) [Ca²⁺]_i time series from beta cells indicated in panel (C), around the interval of one selected burst (middle panel). The triangles indicate the onsets of individual burst, whereby activation time is color‐coded with a gradient scale (dark blue denotes burst onset and light blue 0.24 s afterward). The upper panel indicates the corresponding temporal evolution of coactivity between spikes from different cells as a function of the intercellular distance within a sliding temporal window ΔT _w = 0.3 s and step size ΔT _s = 0.15 s. The lowermost panels show three spike‐based coactivity matrices from different intervals, as indicated by the arrows (T ₁ = 17.25 s; T ₂ = 17.8 s; T ₃ = 18.25 s). (B) Zoom‐in from panel (A), depicting individual spikes with light blue lines, and color‐coded triangles indicate the onsets of individual spikes (dark blue reflects zero and light blue 0.036 s delay from the onset of the first spike in the particular spike‐based event). (C) An islet of Langerhans, loaded with Calbryte 520 AM, during high temporal resolution line scan. Broken line indicates the position of the line scan, and yellow squares indicate ROIs that correspond to individual beta cells. Scale bar 50 μm. Numbers correspond to beta cells in panel (B). (D) Functional connectivity networks extracted from coactivity between spikes (upper panel) and bursts (lower panel). We used a variable threshold (0.5 and 0.875 for spikes and bursts, respectively) to account for differences in the relative overall activity between spikes and bursts. (E) Evolution of average coactivity (color‐coded) derived from spikes along the course of a burst as a function of intercellular distance for 8 mM (left) and 12 mM (right) glucose stimulation. (F) Average coactivity of spikes and bursts as a function of the intercellular distance during stimulation with 8 mM (left) and 12 mM (right) glucose. Shown are means (colored line) and SEM (colored area). (G) The relationship between the average delays of spikes (left) and bursts (right), and distance between cells for 8 mM (blue) and 12 mM (red) glucose. Shown are means (colored line) and SEM (colored area). The results presented in panels (E–G) were obtained on data pooled from the following number of cells/islets: 18/3 (8 mM glucose), 22/3 (12 mM glucose). Only recordings with ≥5 cells were included in the analysis.

We extended the analysis to pooling data from several bursts and islets, separately for the two glucose concentrations (Figure 5E). In both glucose concentrations, the coactivity was a decreasing function of intercellular distance, and it exhibited a bell‐shaped dependence on the delay from the burst onset, with the highest values roughly between 0.3 and 0.9 s after the burst onset. The coactivity in 8 mM glucose was lower than in 12 mM glucose. To examine this issue in more detail, we computed the average coactivity as a function of the Euclidean distance between cell pairs, separately for bursts and spikes (Figure 5F). The average coactivity was a monotonically decreasing function of the intercellular distance, irrespective of the stimulatory glucose concentration and type of activity analyzed. However, there were notable differences in the absolute values as well as in the steepness of this relation. First, the coactivity level between the bursts was higher and deteriorated slower with distance in 12 mM than 8 mM glucose. This reflects the fact that under higher glucose concentrations there is a higher fraction of global [Ca²⁺]_i waves, which is also in agreement with our previous findings in [Ca²⁺]_i imaging studies. ¹⁹ , ⁴² , ⁴³ Second, the level of coactivity between [Ca²⁺]_i spikes was significantly lower than between bursts for both glucose concentrations and it also declined faster with increasing intercellular distance between cell pairs. This observation corroborated the notion that the propagation of [Ca²⁺]_i spikes is often limited to proximal cells only. Moreover, the coactivity between [Ca²⁺]_i spikes seemed to be glucose‐dependent. While for the most nearby cells the coactivity coefficient value was very similar for both glucose concentrations (around 0.6), the degree of aligned [Ca²⁺]_i spikes decreased more rapidly with increasing intercellular distances in 8 mM than in 12 mM glucose. Additionally, we quantified the intercellular activity via functional connectivity networks that were extracted from coactivity matrices computed for the whole recording, separately for the bursts and for the [Ca²⁺]_i spikes. To this purpose, both coactivity matrices were thresholded, whereby the threshold values were adjusted to account for the differences in overall activity in the oscillatory components. Both networks are presented in Figure 5D and it can be observed that the spike‐derived network was more sparse with connections mostly between neighboring cells only. In contrast, the burst‐derived network was denser and more interconnected. The latter reflects the fact that bursts usually travel across all the cells, in contrast to [Ca²⁺]_i spikes, where the transmission is more often limited to adjacent cells.

Next, we assessed whether [Ca²⁺]_i spikes and bursts displayed a spatiotemporal ordered appearance, for example, waves. Detailed analysis demonstrated that the [Ca²⁺]_i bursts form ordered spatiotemporal waves (Figure 5A, color‐coded triangles), well in agreement with previous studies based on lower temporal resolution sampling. ¹⁶ , ¹⁸ , ²⁸ However, looking at a particular burst of [Ca²⁺]_i spikes, the patterns of [Ca²⁺]_i spikes seem to be more heterogeneous and less coherent when compared to the propagation of bursts (Figure 5B). This is at least in part related with the fact that the emergence of spikes does not occur in all cells simultaneously. To quantify the velocity of spike propagation, we quantitatively described the delays between subsequent cellular activations, that is, time lags along individual intercellular waves. In Figure 5G, we show the relation between the average delays between cell pairs and the corresponding intercellular distances. An increasing trend is obtained for bursts and spikes; for bursts, the trend is steeper for 12 mM than for 8 mM glucose whereas for spikes the difference does not appear significant. Apparently, only the velocity of burst propagation seems to be glucose‐dependent. Importantly, it should be noted that the line scan technique does not allow for a direct evaluation of the absolute velocities of intercellular signals in tissue slices, as it assesses the propagation of a three‐dimensional wave in a one‐dimensional line of cells and thus overestimates the velocities. However, by assessing the time lags between subsequent cellular activations it can account for a comparison in relative terms, which indicates that the delays between spikes are approximately 5 times shorter than between bursts. Accordingly, the velocity of propagating spikes seems to be approximately 5 times higher than the velocity of waves that drive the propagation of bursts across the islets.

2.5. The effect of acetylcholine and tetraethylammonium on [Ca²⁺]_i spikes

To gain further insight into the spiking behavior and compare our findings with previous electrophysiological studies, we next studied the effect of acetylcholine (ACh) on [Ca²⁺]_i spikes (Figure 6). Consistent with previous reports, addition of ACh at a physiological concentration (50 nM) to stimulatory glucose (12 mM) elicited oscillations in [Ca²⁺]_i that were qualitatively similar to glucose only (Figure 6A). ⁴⁴ After a stable response to ACh was achieved, sampling [Ca²⁺]_i at a higher rate revealed that individual fast oscillations were populated with spikes (Figure 6B). Binarization and quantification of data indicated that, for spikes, neither the frequency nor the duration or active time differed considerably after addition of ACh (Figure 6C–E). Importantly, ACh significantly affected the bursts: the frequency increased (Q1/M/Q3 0.09/0.12/0.15 Hz in glucose and 0.15/0.19/0.23 Hz after addition of ACh) while the duration decreased (2.2/3.3/5.3 s in glucose and 1.5/1.7/2.5 s after addition of ACh), resulting in a small decrease in active time (0.30/0.38/0.48 for glucose and 0.27/0.32/0.37 after addition of ACh) (Figure 6F6C–G).

FIGURE 6.

Effect of ACh stimulation on [Ca²⁺]_i spiking and bursting activity. (A) [Ca²⁺]_i dynamics during addition of 50 nM ACh to 12 mM glucose. Protocol is indicated at the top of the panel (number indicates glucose concentration in mM). Sampling rate 2 Hz. (B) High sampling [Ca²⁺]_i imaging (61 Hz) of beta cells stimulated with 12 mM glucose + 50 nM ACh. (C–H) Spike (C) and burst (F) frequency, spike (D) and burst (G) duration, and spike (E) and burst (H) active time, for stimulation with 12 mM glucose (blue) and after addition of 50 nM ACh (green). Pooled data from 6356 spikes and 777 bursts from 78 beta cells from 6 islets from 2 mice. Shown are means and SEM for spikes and quartiles for bursts, data for spikes were divided into 7 bins. ***p < 0.001 (Mann–Whitney test).

Further, we tested the effect of addition of tetraethylammonium (TEA) at a low concentration (1 mM) to stimulatory glucose (12 mM) on [Ca²⁺]_i spikes (Figure 7). At this concentration, addition of TEA elicited a qualitatively different bursting pattern, characterized by shortened and more frequent bursts composed of large amplitude spikes (Figure 7A). Occasionally, solitary spikes starting from the baseline and returning to it were observed. Analysis of binarized data confirmed these observations (Figure 7B–H). The duration of bursts was much shorter (Q1/M/Q3 0.45/0.6/1.0 s compared to 2.9/7.2/10.3 s in glucose only) and the frequency much higher (Q1/M/Q3 0.3/0.4/0.6 Hz compared to 0.04/0.05/0.06 Hz in glucose only), resulting in a small decrease in active time (Q1/M/Q3 0.19/0.29/0.39 vs. 0.20/0.34/0.48 in glucose only). For spikes, the duration increased to 140 ms, whereas the frequency and active time were comparable to glucose only. To quantify the effect of TEA on the spike amplitude, we measured spike amplitudes relative to the baseline‐to‐peak amplitude. TEA increased the spike amplitude by almost 100% on average (Q1/M/Q3 28/35/42% vs. 14/18/24% in glucose only).

FIGURE 7.

Effect of TEA stimulation on [Ca²⁺]_i spiking and bursting activity. (A) [Ca²⁺]_i dynamics from beta cells during stimulation with 1 mM tetraethylammonium (TEA) in 12 mM glucose. Inset A1 illustrates the methodology used for amplitude determination. (B–H) Spike (B–D) and burst (E–G) frequency, duration active time, during stimulation with 12 mM glucose (blue) and after addition of 1 mM TEA (magenta). Pooled data from 10 939 spikes and 1737 bursts from 65 beta cells from 6 islets from 2 mice. Shown are means and SEM for spikes and quartiles for bursts, data for spikes were divided into 30 bins. ***p < 0.001, **p < 0.01 (Mann–Whitney test).

2.6. The effect of sampling frequency on analysis of [Ca²⁺]_i data

The sampling frequency of [Ca²⁺]_i imaging is limited in many studies, and this may critically affect the quantification of [Ca²⁺]_i activity, especially the estimation of [Ca²⁺]_i spike and burst shape and duration. We took advantage of the available high temporal resolution [Ca²⁺]_i data to quantitatively study the effect of lower sampling frequencies. As a proof of concept, we selected a beta cell glucose‐induced bursting activity that was clearly resolving the [Ca²⁺]_i spikes and used computer‐based down‐sampling to mimic the effect of lower sampling frequency on the qualitative and quantitative description of [Ca²⁺]_i activity (Figure 8). The extent of down‐sampling, that is, the computed effect of how the signal would look like if lower sampling rates were used, was arbitrarily selected to cover sampling rates in previous studies. For the spikes, ≤50 Hz sampling already distorted their shape, as expected from the inter‐sampling interval of ≥20 milliseconds in this case. For the bursts, sampling at lower frequencies prolonged duration by 13% (±4%) at 2 Hz, 21% (±12%) at 1 Hz, and >100% (112 ± 80%) at 0.5 Hz.

FIGURE 8.

Sampling frequency limits accuracy during [Ca²⁺]_i recordings. (A) A representative burst of [Ca²⁺]_i activity, sampled at a sampling frequency of 178 Hz (top trace). The panels present computer‐based down‐sampling of the signal to selected frequency (Fsample) which was used to determine the [Ca²⁺]_i burst duration. Blue lines indicate the duration at a given sampling frequency. Data before down‐sampling are plotted in the background (gray). (B) The effect of sampling frequency on estimated [Ca²⁺]_i burst duration at half of the amplitude. Plotted are individual values (gray dots) as well as mean values (black lines) and standard errors (black error bars), relative to the duration at the highest sampling frequency. Data polled from 5 bursts from a single beta cell, stimulated with 12 mM glucose.

3. DISCUSSION

The pattern of glucose‐induced activity in beta cells is highly complex, resulting from an interplay of a variety of ion channels and calcium ion fluxes. ⁸ , ⁴⁵ , ⁴⁶ For physiological concentrations of glucose, this pattern consists of at least two temporal domains: a slower oscillatory component composed of smaller depolarizations lasting for a few seconds, and a superimposed faster oscillatory component composed of larger depolarizations lasting for about 100 milliseconds. ⁵ , ⁷ , ⁹ , ⁴⁵ The faster component is commonly termed spikes and the slower bursts (see introduction for other terms used in different studies).

There is a scarcity of data that would demonstrate whether the spikes translate into changes in intracellular calcium concentration ([Ca²⁺]_i). Indices for a [Ca²⁺]_i correlate of the spikes have been derived from patch‐clamp experiments demonstrating that amplitude was affected by the extracellular [Ca²⁺]_i, ⁷ , ⁴⁷ indicating that the spikes are mediated by Ca²⁺ currents. Fridlyand et al. showed that [Ca²⁺]_i spikes are synced with the membrane potential (MP) spikes; however, they did not specify whether the recorded activity originated from a single beta cell or was averaged over an islet; moreover, they did not qualitatively or quantitatively describe the spikes. ³⁵ Langlhofer et al. reported that short membrane depolarization in INS‐1 cells is able to trigger [Ca²⁺]_i changes in the submembrane domain that lasts for about 100 milliseconds. ³⁷ The kinetics of these oscillations strongly resemble the spikes reported in the electrophysiological data. Furthermore, the authors showed that supraphysiological (15 mM) glucose concentration, in about half of cells, triggered similar [Ca²⁺]_i spikes and that these are organized into a burst‐like pattern. Although this pioneering study provided helpful insight, several open questions remain, namely: (i) what is the relationship between the spikes in membrane potential and [Ca²⁺]_i oscillations during stimulation with physiological concentrations of glucose, (ii) do the submembrane changes translate into global [Ca²⁺]_i changes, (iii) do spikes have a role in collective beta cell behavior, and (iv) do they have a role in differentiating glucose levels by beta cells.

3.1. [Ca²⁺]_i spikes closely follow membrane potential spikes

To provide a direct experimental link between the two modalities (MP and [Ca²⁺]_i), we measured the MP activity from a beta cell while recording [Ca²⁺]_i dynamics from its adjacent neighbors. The glucose‐induced pattern of activity was qualitatively identical: a rapid depolarization coincided with an increase in [Ca²⁺]_i to a plateau with a superimposed spiking activity that gradually waned during repolarization and decrease in [Ca²⁺]_i to the pre‐burst level. Although direct comparison would be possible only if MP and [Ca²⁺]_i were recorded from the same cell (technically not feasible in this study due to dilution of the deesterified dye into the pipette via dialysis), this is to the best of our knowledge first experimental evidence directly linking the spiking electrical activity with [Ca²⁺]_i changes during glucose stimulation. Further quantification revealed that the frequency of the spikes gradually decreases from ~10 Hz at burst initiation, in good agreement with the electrophysiological data, ⁸ , ⁴⁵ although slower spiking frequency was also reported. ⁹ Furthermore, the frequency of the MP and [Ca²⁺]_i spikes along a burst is practically the same. The sampling frequency was sufficient to reliably determine the duration of individual spikes, yielding interesting differences between the two signals: MP spikes were about a half of the [Ca²⁺]_i spike duration, irrespective of the relative timing within a given burst. While [Ca²⁺]_i oscillations with kinetics similar to action potential kinetics were demonstrated in other tissues using Calbryte 520 (e.g., mouse sinoatrial node ⁴⁸ ), we cannot exclude the possibility that the temporal difference could at least partly be attributed to insufficient calcium dye kinetics. The alternative explanation is that the [Ca²⁺]_i kinetics is importantly determined by slower mechanisms of Ca²⁺ removal. Simultaneous measurement of both signals from the same cell using a potassium salt form of the dye or parallel imaging with voltage‐sensitive dyes will be needed to confirm this and, further, determine the exact relative timing.

3.2. Mechanisms governing the [Ca²⁺]_i spike generation

There is still no general consensus in the field regarding the mechanism behind the complex pattern of the [Ca²⁺]_i spikes. While Ca²⁺ ions undoubtedly carry the majority of the currents that generate the spikes, ⁷ , ⁸ , ⁴⁷ , ⁴⁹ several hypotheses have been put forward regarding the mechanistic substrate for the shape of spikes and for the gradual waning of spikes along a burst. According to one hypothesis, ⁴⁹ a glucose‐induced decrease in K_ATP conductance leads to gradual depolarization above the threshold for voltage‐gated calcium channels (VACCs), and their activation is responsible for the fast upstroke of a spike. The downstroke of a spike, on the other hand, is due to voltage‐activated potassium conductance (K_v), and its slower kinetics leads to the asymmetrical shape of a spike. The spike asymmetry has been extensively demonstrated in electrophysiological recordings, and our data corroborated the asymmetry also for [Ca²⁺]_i (compare Figure 4). The buildup of [Ca²⁺]_i during a burst, due to VACC activity, was suggested to cause the gradual waning of the spiking activity via activation of calcium‐induced potassium conductance (K_Ca), leading to slow repolarization and eventually burst termination. ⁴⁹ The effects of TEA in stimulatory glucose in our study corroborate the role of K_Ca (Figure 7). TEA was shown to block K_v (IC₅₀ of 1.4 mM, ⁵⁰ ) when applied at concentrations as low as in our study, and in our hands, both the amplitude and duration of [Ca²⁺]_i spikes increased, presumably due a to TEA‐induced decrease in the velocity of repolarization (Figure 7). Additionally, a greater activation of K_Ca due to increased [Ca²⁺]_i may explain the shortened bursting seen in our experiments, exemplified by occasional single large [Ca²⁺]_i spikes. Stimulation with glucose only, on the other hand, likely involves different mechanisms. Namely, the aforementioned mechanism of burst generation implies a gradual MP repolarization and an increase in [Ca²⁺]_i along the burst. While there are extensive data supporting the MP repolarization during a burst, ⁵ , ⁹ , ⁴⁵ , ⁵¹ , ⁵² our experimental data do not speak for a [Ca²⁺]_i buildup. More specifically, we demonstrated that the [Ca²⁺]_i levels decrease or remain constant during the burst rather than increase (compare Figure 2), suggesting that other mechanisms might be involved. Possibly, [Ca²⁺]_i feedback on ATP/ADP levels might also play a role. ⁵³ An alternative explanation ⁸ , ⁵⁴ that agrees with our experimental observations suggests that VACCs display two inactivation time courses: a faster and a slower component. While the faster component would, in addition to K_v, lead the downstroke of a [Ca²⁺]_i spike, the slower inactivation would mediate the slow repolarization during a burst, leading to a gradually lower open probability of VACCs, a decrease in [Ca²⁺]_i along the burst, and ultimately burst termination. Finally, we explored the influence of ACh on the pattern of spiking and bursting to compare our findings with previous studies employing ACh and explore whether the effect of ACh on insulin secretion can at least partly be accounted for by the effect on the spiking and bursting behavior. Addition of a low concentration ACh, which is known to act on beta cells via M₃ receptors, had practically no effect on either spike frequency, duration, or active time (Figure 6). Interestingly, the bursting pattern changed considerably and in agreement with previous studies. ⁵⁵ More specifically, adding ACh shortened the bursts and made them more frequent, resulting in an overall comparable active time, as reported previously. ⁴⁴ Keeping in mind the positive effect of parasympathetic stimulation on insulin secretion, ⁵⁶ the largely unaltered active time of bursts might appear counterintuitive, given that increasing glucose concentrations increase insulin secretion at least to an important extent due to increases in active time. ¹² , ¹⁹ , ²⁶ Considering that spike active time was the highest at the beginning of bursts, the greater number of shorter bursts during ACh stimulation would generate more spikes cumulatively, suggesting that the total number of spikes per time may be the most important parameter, which is able to account for the effects of both increasing concentrations of glucose, ACh, and possibly other secretagogues. Additionally, ACh has been shown to potently increase the sensitivity of the exocytotic machinery to [Ca²⁺]_i and the relative contribution of each of the mechanisms at different concentrations of ACh will need to be assessed in the future. ⁵⁶ In any case, our findings with ACh are in line with previous findings on the effects of ACh at low doses during the plateau phase, where ACh is not expected to exert any strong effects via Ca²⁺ release from the endoplasmic reticulum, but rather via cell depolarization. ⁵⁶ , ⁵⁷ further studies using more specific agonists and antagonists will be needed to explore the mechanisms of spike generation into more details.

3.3. Role of [Ca²⁺]_i spikes in response to increasing glucose concentrations

Insulin secretion in proportion with the glucose load is of paramount importance to maintain normoglycemia, and several data demonstrate that beta cells are capable of transducing the glucose concentration into electrical and [Ca]_i activity, as well as insulin secretion rate. What is the role of [Ca²⁺]_i spikes in coding the glucose concentration? Most of the data are available for bursts, involving modulation of either their frequency, duration, or both. ⁷ , ¹⁰ , ¹² , ¹⁹ In this study, the higher burst frequency and active time are in accordance with previous studies. ⁷ , ⁹ , ¹² , ¹⁹ For the spikes, on the other hand, only a few studies addressed a possible modulation of MP spike frequency. ⁸ , ⁹ In our hands, [Ca²⁺]_i spike frequency was lower in higher glucose concentration, in contrast to previous studies. However, the spike duration was significantly longer in higher glucose and as a result, the relative active time of spiking activity was higher by about 10%. There are no experimental data that would link the modulation of [Ca²⁺]_i spikes with insulin secretion. However, based on experimental evidence from electrophysiological recordings using trains of depolarizations to mimic spikes and showing that such spikes are very efficient in eliciting granule fusion, it is reasonable to speculate that spikes are the main trigger for secretion and that the duration of bursts determines the total number of spikes and thus possible secretory events. If we accept this proposition, then the currently standard method of estimating active time from bursts overestimates the “true” active time by a factor of 2–4, since the spike active time within a burst is between 25 and 50% in most cases. Moreover, even if the frequency of spikes was less in 12 mM glucose, the active time of bursts was much longer and thus the total number of spikes during activity significantly higher. To what extent the duration of spikes and their active time play a role, needs to be determined in the future by using a higher number of different glucose concentrations and employing additional studies with trains mimicking different recorded spikes and measuring the accompanying capacitance changes. Currently, we can conclude that the modulation of burst active time seems to contribute more, since the spike active time does not seem to be strongly affected by the concentration of glucose, however, this mechanism may prove to be crucial for some other secretagogues, such as ACh (see above).

3.4. Spatiotemporal characteristics of [Ca²⁺]_i spikes

Gap‐junctional coupling between beta cells strongly shapes their coordinated activity within an islet, a property that has recently gained much attention in the field of beta cell research. ³¹ , ⁵⁸ , ⁵⁹ , ⁶⁰ For the [Ca²⁺]_i bursts, several groups have demonstrated characteristic spatiotemporal organization in the form of repetitive waves spreading over an islet with a speed of about 100 μm/s. ¹⁶ , ¹⁸ , ²⁸ , ⁴² For the spikes, data are limited to single or two‐cell MP recordings, demonstrating synchronous MP spikes in nearby cells. ¹⁴ , ⁵² We demonstrated that the coordination of [Ca²⁺]_i spikes was limited by the extent of overlapping beta cell activity during bursts, resulting in a heterogeneous overall coactivity due to non‐overlapping periods of the bursts. Occasional synchronous global [Ca²⁺]_i spikes were detected, and the increasing trend between average time delays between cell pairs and the corresponding intercellular distances implies that both spikes and bursts are transmitted between cells by means of intercellular waves. Our data further suggest different velocities for bursts at a velocity of about 100 μm/s ²⁸ and an estimated 5 times faster propagation for [Ca²⁺]_i spikes. This is in accordance with the cable theory predicting that the depolarization velocity is higher for larger and faster depolarization rates, which is certainly the case for spikes compared with bursts. ⁶¹ Noteworthy, the estimated velocity of spike waves in this study is in good agreement with our previous finding on depolarization and [Ca²⁺] wave velocity during stimulation with glucose and tetraethylammonium at high concentration. ²² Further, the burst wave velocity and burst duration are still high enough to ensure that all beta cells within an islet are coactive for a couple of seconds, which suggests that it may be important that they secrete insulin together during this time. Comparing the spiking and bursting behavior with data on insulin secretion, it has been shown that during the plateau phase, beta cells secrete at a rate of approximately 5 granules per min, which means roughly one granule per burst or less. ⁶² Thus, it seems tempting to speculate that in most cases, insulin secretion is initiated by a single spike within a burst, but which part of a burst is responsible for granule fusion and whether the synchronization of spikes is important for normal insulin secretion will have to be determined in the future. Finally, spikes in an already active cell are unable to trigger spiking in the neighboring cells, as long as their baseline MP is below the threshold for burst initiation; however, electrical currents directed from the active cell into the inactive neighbor aid to its depolarization and facilitate burst initiation. This is also the reason why the nature of multicellular activity differs when comparing spikes and bursts. After the initial and transient period upon switching from substimulatory to stimulatory conditions, the bursts become rather coherent. In this so‐called plateau phase of sustained activity, intercellular waves characterizing burst propagation often involve most cells. As a result, the functional networks derived from coactivity are rather dense and exhibit also long‐range connections (Figure 5D, lower panel). In contrast, the propagation of spiking activity is typically confined to a few neighboring cells, as greater spike propagation occurs only during the intermediate period when burst peaks overlap. Consequently, coactivity‐based networks derived from spiking activity exhibit sparser connections, predominantly between proximate cells only (Figure 5D, upper panel). Notably, this observation is consistent with theoretical predictions, which have demonstrated that the nature of functional networks in electrically coupled syncytia is shaped by the characteristics of intercellular waves. ⁶³ Expanding the scope of our study, future investigations into coordinated [Ca²⁺]_i spiking should focus on pressing topics such as the relations to hub, wave initiator, and first responder cells. However, current experimental capabilities allow for high‐resolution ultrafast recordings primarily at the level of line scans. Studies aiming to address how specific subpopulations of cells shape multicellular spiking activity will need to await advancements in technical capabilities that enable high‐resolution imaging across an entire tissue slice.

3.5. The required minimal sampling rate: guidance for future experiments

The unprecedented temporal resolution in this study provided a unique opportunity to qualitatively reconcile previous reports on beta cell activity utilizing slower sampling frequencies (≤5 Hz). ²¹ , ²³ , ²⁵ , ³² We demonstrated that sampling rates of >20 Hz generate a negligible overestimation of burst durations, sampling at 2 Hz an overestimation of about 10% and that the errors in estimating burst duration are then progressively increasing for lower sampling frequencies, with sampling at 0.5 Hz resulting in an overestimation by 100% (Figure 8). It is important to keep these limitations in mind when comparing past studies reporting [Ca²⁺]_i oscillations with different sampling frequencies or when studying subtle effects of tested substances on burst duration.

4. METHODS

4.1. Ethics statement

We conducted the study in strict accordance with all national and European recommendations on care and handling experimental animals, and all efforts were made to minimize the suffering of animals. The Administration of the Republic of Slovenia for Food Safety, Veterinary and Plant Protection approved the experimental protocol (permit numbers: U34401‐35/2018‐2).

4.2. Animals and pancreas tissue slice preparation

Three to six month‐old male NMRI mice were housed on a 12:12 hours light: dark schedule in individually ventilated cages (Allentown LLC, USA). Preparation of acute pancreas tissue slices was described previously. ¹⁸ , ⁶⁴ In brief, after sacrificing the animal, the abdominal cavity was accessed via laparotomy and we distally clamped the common bile duct at the major duodenal papilla. The low‐melting‐point 1.9% agarose (Lonza, USA) dissolved in extracellular solution (ECS, consisting of (in mM) 125 NaCl, 26 NaHCO₃, 6 glucose, 6 lactic acid, 3 myo‐inositol, 2.5 KCl, 2 Na‐pyruvate, 2 CaCl₂, 1.25 NaH₂PO₄, 1 MgCl₂, 0.5 ascorbic acid) at 37–40°C was injected into the common bile duct. Following pancreas extraction in ice‐cold ECS, we prepared 140 μm thick tissue slices with a vibratome (VT 1000 S, Leica). HEPES‐buffered saline at RT (HBS, consisting of (in mM) 150 NaCl, 10 HEPES, 6 glucose, 5 KCl, 2 CaCl₂, 1 MgCl₂; titrated to pH = 7.4 using 1 M NaOH) was used to collect the slices. For staining, we incubated the slices for 50 minutes at RT in the dye‐loading solution (6 μM Calbryte 520 AM (AAT Bioquest), 0.03% Pluronic F‐127 (w/v), and 0.12% dimethyl sulfoxide (v/v) dissolved in HBS). Slices were stored in fresh HBS at RT until experimentation. All chemicals were obtained from Sigma‐Aldrich (St. Louis, Missouri, USA) unless otherwise specified.

4.3. [Ca²⁺]_i imaging

For [Ca²⁺]_i imaging, we transferred an individual slice into a recording chamber of Leica TCS SP5 AOBS Tandem II upright confocal system (20× HCX APO L water immersion objective, NA 1.0) or LEICA SP8 Stellaris upright confocal system (25× HC IRAPO water immersion objective, NA 1.0) (Leica Microsystems, Germany). Slices were continuously perifused with carbogenated ECS at 37°C. We perifused the slices with the ECS containing 6 mM glucose during setting up of recording parameters. In the majority of experiments, these were initially set to 1–2 Hz at 512 × 512 pixels and then switched to a higher sampling rate line scan for 1–3 min. The lower sampling settings were also applied during switching of perifusion to ECS containing glucose, acetylcholine (ACh), or tetraethylammonium (TEA) until a stable beta cell response was observed, and then reverted to the higher sampling setting for 1–3 min. To resolve the ultrafast [Ca²⁺]_i dynamics, we recorded a line scan at 8000 or 12 000 Hz using line accumulation of 16. Depending on the intensity of fluorescence staining, we adjusted the recording parameters. If intensity was higher, setting to 512 pixels with a frame size of 8192 lines per frame enabled a 1498 Hz sampling rate. If the intensity was lower, we used a lower sampling rate to achieve sufficient signal‐to‐noise ratio (e.g., 1024 pixels with 2 lines per frame and additional line average of 32 achieved 165 Hz). During off‐line analysis, data were smoothed as described in data processing and analysis section. In a subset of experiments aimed at assessing the 1^st phase of the response, the fast scanning started before stimulation with glucose, and due to prolonged duration (about 10 min), the fluorophore excitation and therefore photobleaching was reduced, resulting in an average sampling rate of about 50 Hz. Immediately before or after the line scan, an additional high‐resolution reference image (512 × 512 or 1024 × 1024 pixels) was acquired. The calcium reporter dye was excited by a 488 nm argon laser line, and the emitted fluorescence collected with HyD or HyD S detectors in photon counting mode in the range of 500–700 nm.

4.4. Membrane potential recording in combination with [Ca²⁺]_i imaging

We simultaneously combined recording of [Ca²⁺]_i as described in the previous section with a whole‐cell patch‐clamp approach. ²² In brief, we used standard whole‐cell mode using patch‐clamp lock‐in amplifier (SWAM IIc, Celica, Slovenia) and 2–3 MΩ borosilicate glass capillaries filled with (in mM) 125 potassium methanesulfonate, 20 KCl, 40 HEPES, 2 MgCl₂, 5 Na₂ATP, titrated to pH = 7.2 using 1 M KOH, osmolarity 300 ± 10 mOsm. The patch‐clamp setup was controlled with WinWCP (Strathclyde Electrophysiology Software, John Dempster, University of Strathclyde, Glasgow, UK) and coupled to the LASAF software (electrophysiology plug‐in) via a TTL trigger. Sampling frequency of the membrane potential recording was set at 10 kH. A two‐expert‐person coordination was needed to simultaneously control both electrophysiological and [Ca²⁺]_i imaging system.

4.5. Data processing and analysis

Line scans were exported off‐line from LASAF software (Leica Microsystems, Germany) as sets of images together with corresponding time stamps, and membrane potential data were exported from the WinWCP as text files. Further analysis was performed with in‐house MATLAB/Phyton scripts. For [Ca²⁺]_i data, we manually selected ROIs that corresponded to individual beta cells using three criteria. First, identity of beta cells was assessed based on the activity pattern: a lack of activity during substimulatory 6 mM ECS perifusion, and transient 1^st phase response followed by repetitive and coordinated oscillation during stimulatory 8 or 12 mM ECS perifusion, typical for beta cells. ¹⁸ , ⁶⁵ , ⁶⁶ , ⁶⁷ Further, to discriminate individual beta cells in a line scan, we resorted to morphological features on the reference high‐resolution images that displayed cell boundaries and differences in temporal activity such as delays in [Ca²⁺]_i oscillations between cells.

Data from ROI selection and membrane potential data with time tags were exported as time series for further analysis. Activity was binarized by extracting (i) the [Ca²⁺]_i spikes and (ii) the underlying shape of the [Ca²⁺]_i change on which the spikes were superimposed. Data were smoothed with local regression function using weighted linear least squares and a 2^nd degree polynomial model with a moving average of 10–50 data points, assuring that the smoothing did not affect the shape of [Ca²⁺]_i spikes nor introducing any time shift. Larger windows for smoothing were used with higher sampling rates and lower values with lower sampling rates. Such processing represents a well‐established approach for evaluating rapid cellular oscillations, such as neuronal firing, and was shown to enhance the accuracy of subsequent analyses. ⁶⁸ We first extracted the underlying shape of bursts by applying a rank order filter with rank = 50^th percentile and a sliding window of 150–200 data points, with these parameters empirically selected to ensure optimal precision. This filter faithfully eliminated the [Ca²⁺]_i spiking activity and extracted the underlying profile of [Ca²⁺]_i change, allowing for burst binarization defined as intervals at half‐maximal amplitude. In the second step, the [Ca²⁺]_i spiking activity was extracted by subtraction of the filtered signal from the raw data and binarized as time points at half‐maximal amplitude. Notably, similar procedures are commonly employed to separate distinct oscillatory scales in multimodal cellular and physiological signals. ⁶⁹ Finally, binarized data of either bursts or spikes were used to calculate durations, frequencies (derived from inter‐event time intervals), and the active time (i.e., percentage of time occupied by the spikes or bursts), as described previously. ¹⁸ , ¹⁹ , ²² , ²⁶ Data were reported either as the mean and SEM or as the median and quartiles. To compare temporal evolution of the [Ca²⁺]_i spikes from different bursts, we plotted data relative to burst onset.

4.6. Quantifying intercellular activity

To assess intercellular interactions and synchronicity, we computed the coactivity coefficients for the binarized dynamics of [Ca²⁺]_i bursts and spikes. Specifically, the ij‐th element in the coactivity matrix was defined as $C_{\mathit{ij}}=T_{\mathit{ij}}/\sqrt{T_i{T}_j}$ and reflected the degree of synchronous activity between the i‐th and the j‐th cell. Here $T_{\mathit{ij}}$ stand for the total coactivity time in which both cells were simultaneously active and $T_i$ and $T_j$ are the total activity time for both cells i and j. Coactivity is a well‐established method for evaluating synchronous cellular activity, particularly in the field of neuroscience, ⁷⁰ and has also proven effective for quantifying the coordination of beta cell dynamics. ⁴¹ , ⁷¹ To characterize the temporal evolution of coactivity of [Ca²⁺]_i spike during a burst, we computed the series of coactivity matrices using a sliding window of ΔT _w = 0.3 s and step ΔT _s = 0.15 s. To compute the average coactivity as a function of intercellular distance, we averaged over all cell pairs within the given distance interval: $\mathrm{CA}\left(d_{\mathit{ij}}\right)=\frac{1}{n}{\sum }_{\mathrm{\Omega }}{C}_{\mathit{ij}}$, where Ω denotes the subset of cells positioned within the given distance interval, n is the number of cells in this interval, and $\mathrm{CA}\left(d_{\mathit{ij}}\right)$ denotes the average coactivity at the given intercellular distance $d_{\mathit{ij}}$. For the characterization of the intercellular signal propagation, we used the space‐time cluster algorithm, as described previously. ⁷¹ , ⁷² In brief, the binary traces of individual cells (bursts or spikes) along with their physical positions were translated to a time‐space cube. Within this cube, cells that exhibited concurrent activity within a narrow timeframe were identified as part of the same intercellular Ca²⁺ event. Within a given event referring either to [Ca²⁺]_i bursts or [Ca²⁺]_i spikes, the cellular activation sequence and the corresponding delays in activations between cell pairs were extracted.

5. CONCLUSION

In conclusion, high‐frequency calcium imaging is currently technically demanding and limited to line scans. Still, it may be worth the effort for specific scientific questions. With the advent of even better and more affordable technological solutions and in combination with specific pharmacological, genetic, or dietary interventions, it shall help researchers in the field of islet physiology to detect novel mechanisms of normal and pathological beta cell functioning.

AUTHOR CONTRIBUTIONS

Jurij Dolenšek: Conceptualization; methodology; software; data curation; supervision; validation; investigation; funding acquisition; visualization; resources; writing – original draft; writing – review and editing. Viljem Pohorec: Data curation; writing – review and editing; writing – original draft; investigation; validation; formal analysis; visualization. Maša Skelin Klemen: Data curation; investigation; validation; visualization; writing – original draft; writing – review and editing. Marko Gosak: Software; methodology; formal analysis; resources; visualization; funding acquisition; writing – original draft; writing – review and editing. Andraž Stožer: Conceptualization; methodology; investigation; supervision; funding acquisition; visualization; project administration; resources; writing – original draft; writing – review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Dolenšek J, Pohorec V, Skelin Klemen M, Gosak M, Stožer A. Ultrafast multicellular calcium imaging of calcium spikes in mouse beta cells in tissue slices. Acta Physiol. 2025;241:e14261. doi: 10.1111/apha.14261

DATA AVAILABILITY STATEMENT

Data are available on request from the authors.