Imaging Calcium Dynamics in Subpopulations of Mouse Pancreatic Islet Cells

Summary

Here, we present a protocol for imaging and quantifying calcium dynamics in heterogeneous cell populations, such as pancreatic islet cells. Fluorescent reporters are delivered into the peripheral layer of cells within the islet, which is then immobilized and imaged, and per-cell analysis of the dynamics of fluorescence intensity is performed.

Abstract

Pancreatic islet hormones regulate blood glucose homeostasis. Changes in blood glucose induce oscillations of cytosolic calcium in pancreatic islet cells that trigger secretion of three main hormones: insulin (from β-cells), glucagon (α-cells) and somatostatin (δ-cells). β-Cells, which make up the majority of islet cells and are electrically coupled to each other, respond to the glucose stimulus as one single entity. The excitability of the minor subpopulations, α-cells and δ-cells (making up around 20% (30%) and 4% (10%) of the total rodent¹ (human²) islet cell numbers, respectively) is less predictable and is therefore of special interest.

Calcium sensors are delivered into the peripheral layer of cells within the isolated islet. The islet or a group of islets is then immobilized and imaged using a fluorescence microscope. The choice of the imaging mode is between higher throughput (wide-field) and better spatial resolution (confocal). Conventionally, laser scanning confocal microscopy is used for imaging tissue, as it provides the best separation of the signal between the neighboring cells. A wide-field system can be utilized too, if the contaminating signal from the dominating population of β-cells is minimized.

Once calcium dynamics in response to specific stimuli have been recorded, data are expressed in numerical form as fluorescence intensity vs. time, normalized to the initial fluorescence and baseline-corrected, to remove the effects linked to bleaching of the fluorophore. Changes in the spike frequency or partial area under the curve (pAUC) are computed vs. time, to quantify the observed effects. pAUC is more sensitive and quite robust whereas spiking frequency provides more information on the mechanism of calcium increase.

Minor cell subpopulations can be identified using functional responses to marker compounds, such as adrenaline and ghrelin, that induce changes in cytosolic calcium in a specific populations of islet cells.

Introduction

The purpose of the method is to image real-time changes in cytosolic calcium concentration ([Ca²⁺]_cyt) in minor subpopulations of pancreatic islet cells. This allows uncovering the mechanisms governing hormone secretion in these cells, revealing details about the cross-talk between different cell types and, potentially, introducing a populational dimension into the larger picture of islet signaling.

Islets consist of several cell types. Besides the more well-known insulin-secreting β-cells, there are at least two subpopulations that are also critical in regulating blood glucose³. α-Cells (that make up around 17% of islet cells) secrete glucagon when blood glucose gets too low, which signals for release of glucose into the bloodstream from depots in the liver. Excessive glucagon levels (hyperglucagonemia) and impaired control of glucagon-release accompany (and, technically, can contribute to) the prediabetic condition of impaired insulin sensitivity⁴. δ-Cells (around 2%) secrete somatostatin in response to glucose elevation. This ubiquitous peptide hormone is likely to be present at high concentrations in the vicinity of α- and β-cells within islets, which has a strong G_i receptor-mediated attenuating effect on both glucagon and insulin secretion.

α-Cells and δ-cells share a large part of the glucose-sensing machinery with their close lineage relatives, β-cells. All three cells types are equipped with ATP-sensitive K⁺ channels, elaborate metabolic sensors⁵ that control the plasma membrane potential of these excitable cells. At the same time, secretion of insulin, somatostatin and glucagon is regulated differently by glucose. Imaging of Ca²⁺ dynamics in the two minor subpopulations of islet cells can therefore provide an insight into the cross-talk between blood glucose and islet secretory output.

Early attempts of monitoring the excitability of α- and δ-cells using patch-clamp electrophysiology were soon followed by imaging of Ca²⁺ in single α- and δ-cells. The identity of cells in these experiments was verified via a posteriori staining with anti-glucagon or anti-somatostatin antibodies. These efforts were frequently hampered by the finding that islet cells behave very differently within the islet and as single cells. Although β-cells may appear to be the main benefactors of the islet arrangement (due to their overwhelming majority that underlies their strong electrical coupling), the main discrepancy was, surprisingly, found in α-cells. Within the intact islet, these cells are constantly and persistently activated at low glucose, which is only true for around 7% of single dispersed α-cells⁶. Reporting the activity of α- and δ-cells within intact islets is therefore believed to represent a closer approximation of in vivo conditions.

In general, there are two ways of reporting Ca²⁺ dynamics specifically from the α-cell or δ-cell subpopulations: (i) expressing a genetically encoded Ca²⁺ sensor via a tissue-specific promoter or (ii) using marker compounds. The more elegant former approach adds the substantial advantage of true 3D imaging and hence studying of cell distribution within the islet. It cannot however be applied for intact human islet material. Another potential concern is the 'leakiness' of the promoter, particularly when the β-/α-cell transdifferentiation or α-cell response to high glucose is in place. The latter approach can be used with freshly isolated tissue including human samples or cultured islets. The data, however, is collected solely from the peripheral layer of islet cells, as delivering the dye/marker molecule in deeper layers without altering the islet architecture is challenging. An unexpected advantage of the latter approach is the compatibility with wide-field imaging mode, which allows scaling up the experiments to simultaneous imaging of tens or hundreds of islets (i.e., thousands to tens of thousands of cells).

Calcium is imaged in vivo using genetically encoded GCaMP⁷ (or pericam⁸) family sensors, which are variants of circularly permutated green fluorescent protein (GFP) fused to the calcium-binding protein calmodulin and its target sequence, M13 fragment of myosin light chain kinase^7,9. GCaMPs have superb signal-to-noise ratios in the range of nanomolar Ca²⁺ concentrations and a high 2-photon cross-section, which makes them an ideal choice for in vivo work^10,11. The challenging aspect of using recombinant sensors is their delivery into the cells. Heterologous expression requires using a viral vector and multi-hour ex vivo culturing, which frequently raises concerns regarding potential de-differentiation or deterioration of cell functions. Although mouse models pre-engineered to express GCaMP address this problem, they add new challenges by increasing the lead time substantially and limiting the work to a non-human model. Very high sensitivity to changes of intracellular pH is another adverse side of protein-based sensors¹², which is, however, less of a problem for sensing oscillatory signals, such as Ca²⁺.

The advantage of trappable dyes (such as green fluorescent Fluo4) is that they can be loaded into freshly isolated tissue within around an hour. Predictably, trappable dyes have lower signal-to-noise ratios and (much) lower photostability than their recombinant counterparts. We cannot confirm¹³ the reports of toxicity of the trappable dyes¹⁴, however, dye overloading is a frequent problem.

Red recombinant Ca²⁺ sensors based on circular permutation have been evolving rapidly since 2011¹⁵, and most recent developments present a strong competition to GCaMPs¹⁶ for tissue imaging, given higher depth of penetration of red light. Commercially available red trappable dyes can be used reliably for single-cell imaging but, on the tissue level, cannot compete well with the green analogs.

There is seemingly very little choice of imaging technology for experiments in tissue where out-of-focus light becomes a critical problem. The confocal system provides acceptable single-cell resolution by cancellation of the out-of-focus light with any objective on the NA above 0.3 (for the case of GCaMP6) or 0.8 (trappable dye). In a technical sense, a conventional confocal microscope can be used for simultaneous imaging of [Ca²⁺]_cyt from hundreds (GCaMP) or tens of islets (trappable dye). The only realistic alternative to confocal mode in case of 3D expression of the sensor in tissue is perhaps light-sheet microscopy.

Things are slightly different for the case when the sensor is expressed in the peripheral layer of cells within the islet tissue. For bright recombinant sensors that have a vivid intracellular expression pattern, using a wide-field imaging mode with a low-NA objective may provide sufficient quality and reward the researcher with a substantial increase in the field of view area and hence the throughput. A wide-field system provides poorer spatial resolution, as the out-of-focus light is not cancelled; therefore, imaging tissue with high-NA (low depth of field) objectives is less informative, as the single-cell signal is vastly contaminated by neighboring cells. The contamination is much smaller for low-NA (high depth of field) objectives.

There are tasks, however, for which high throughput and/or sampling rate become a critical advantage. α- and δ-cells exhibit substantial heterogeneity, which creates a demand for high sample sizes to reveal the contribution of the subpopulations. Wide-field imaging is fast and more sensitive, with an industrial-scale large field-of-view system imaging hundreds (GCaMP) or tens (Fluo4) of islets at the same signal-to-noise ratio as the confocal experiments on ten or a single islet, respectively. This difference in throughput makes the wide-field system advantageous for populational imaging with a single-cell resolution, which can be especially critical for small subpopulations such as the δ-cell one. Likewise, attempts to reconstruct electrical activity from Ca²⁺ spiking¹⁷ would benefit from the higher sampling rate provided by a wide-field imaging mode. At the same time, several "niche" problems like the activity of pancreatic α-cells upon stimulation of the dominating β-cell subpopulation, require the use of a confocal system. A factor that influences the decision towards confocal mode is the presence of substantial contaminating signal from the β-cell subpopulation.

Although using hormone-specific antibody staining to verify the identity of the cells after the imaging experiments is still an option, minor cell subpopulations can be identified using functional marker compounds, such as adrenaline and ghrelin that were shown to selectively stimulate Ca²⁺ dynamics in α-¹⁸ and δ-cells^19,20, respectively.

The analysis of time-lapse imaging data aims to provide information beyond trivial pharmacology, such as populational heterogeneity, correlation and interaction of different signals. Conventionally, imaging data is analyzed as intensity vs. time and normalized to the initial fluorescence (F/F₀). Baseline correction is frequently needed, due to the bleaching of the fluorophore signal or contamination by changes in autofluorescence or pH (typically induced by millimolar levels of glucose¹²). Ca²⁺ data can be analyzed in many different ways, but three main trends are to measure changes in the spike frequency, the plateau fraction, or area under the curve, computed vs. time. We found the latter approach advantageous, especially in application to heavily undersampled confocal data. The advantage of the pAUC metric is its sensitivity to both changes in signal frequency and amplitude, whereas computing the frequency requires a substantial number of oscillations²¹, which is hard to attain using conventional imaging. The limiting factor of pAUC analysis is its high sensitivity to baseline changes.

Protocol

All methods described here were developed in accordance with the United Kingdom Animals (Scientific Procedures) Act (1986) and the University of Oxford ethical guidelines.

1. Isolate mouse pancreatic islets

1. Prepare the culture medium and the isolation solution.
   1. Make up the culture medium: RMPI1640 (see Table of Materials), supplemented with 10% of fetal calf serum, 100 units/mL penicillin, 100 μg/mL streptomycin. Dispense the medium into two 60 mm plastic Petri dishes (not treated with any adhesive), and keep the dishes in the incubator (37 °C, 5% CO₂, absolute humidity).
   2. Make up the isolation solution (50 mL/mouse): Hanks' medium with 5 mM glucose, 100 units/mL penicillin 100 μg/mL streptomycin. Keep on ice (4 °C).
   3. Make up the enzyme solution (e.g., Liberase): 0.2 mg/mL in the isolation solution, 2 mL per mouse. Keep on ice (4 °C). Optionally, pre-test the activity of liberase, and optimize the incubation parameters for each batch of the enzyme²².
2. Inject the enzyme solution into the mouse bile duct.
   1. Sacrifice the mouse (12-week old female, C57Bl/6J) via cervical dislocation.
   2. Under a dissection microscope, cut open the skin and muscle layers using fine scissors and locate the bile duct.
   3. Ligate the intestine on both sides of the junction with the bile duct, using a cotton thread.
   4. Lift the bile duct gently with curved fine watchmaker’s forceps and introduce the ice-cold enzyme solution into the duct using a 2 mL syringe with a 30G needle. An inflation of the pancreas should be observed at this stage.
   5. Collect the inflated pancreas using a combination of blunt-nosed thumb forceps and fine watchmaker’s forceps into a 15 mL falcon tube and keep it on ice (4 °C). Add 1 mL of the enzyme solution into the tube.
3. Liberate and collect the pancreatic Islets from the pancreas.
   1. Incubate the pancreata inflated with the enzyme solution for 16 min in a water bath, at 37 °C. Gentle shaking can be used but is not critical, provided the inflation has been good.
   2. Stop the digestion by addition of ice-cold isolation solution, up to 10 mL. Gently shake the digest: it should fall into small pieces.
   3. Wash the digest three times in 10 mL of the isolation solution, let stand 5 min at 1 x g to sediment the liberated islets, on ice. Aspirate the supernatant gently, with a 10 mL serological pipette.
   4. Add cold RPMI to the digest (to make up to 10 mL).
   5. Use plastic Petri dishes (step 1.1.1) to pick the islets. Decant the RMPI medium off one of the Petri dishes, and gently pour some (4-5 mL) of the digest instead. Collect the liberated islets that appear as round, smooth and high-density pieces, with a P10 pipette into the second Petri dish.
   6. Culture the islets in the incubator (37 °C, 5% CO₂, absolute humidity). The experiment can be paused at this stage. Leaving the islets for 1-2 hours is believed by some to help to recover from the mechanical stress during the isolation.

2. Load the dye or express the sensor

1. Prepare the trappable dye.
   1. Dissolve the aliquot (normally, 50 μg) of the trappable dye (e.g., Fluo-4) in DMSO to a stock concentration of 2 mM. Add pluronic acid to a final concentration of 1% (using a 20% stock in DMSO), to improve the solubilization of the dye.
   2. Aliquot the dye in small PCR tubes (2 μL into each). The dye can be stored frozen (-20 °C) for several weeks.
2. Alternatively, prepare the recombinant sensor.
   1. Distribute the sensor (e.g. adenoviral vector encoding GCaMP6f) into 10 μL aliquots and store at -80 °C.
   2. (Optionally) pre-test the titer of the viral stock by infecting islets (as below) using several serial dilutions, to reveal the optimal concentration for infection.
      NOTE: Recombinant sensors encoded by different vectors (lentivirus, BacMam, AAV) require a different infection protocol. Make sure to check this with the vector provider and optimize the working ratio for your needs. The "working" stock of the adenovirus can be stored at -20 °C and frozen/thawed several times. Excessive freeze-thaw cycling reduces the effective titer of the virus.
3. Prepare the imaging solution.
   1. Make up the imaging solution, mM: 140 NaCl, 4.6 KCl, 2.6 CaCl₂, 1.2 MgCl₂, 1 NaH₂PO₄, 5 NaHCO₃, 10 HEPES (pH 7.4, with NaOH).
   2. Make up a stock of glucose (0.5 M) and mannitol (0.5 M) in the imaging solution. The stock can be stored in the fridge (4 °C) for several weeks.
4. Load the trappable dye.
   1. Make up the dye working solution by dissolving 2 μL of the dye in 600 μL of the imaging solution containing 6 mM glucose. The solution can be heated up or vortexed to improve the solubilization.
   2. Load the islets isolated in step 1 into the working solution of the dye. The loading can be done using a multiwell plate or a Petri dish. In the latter case, place a 100 μL droplet of the working solution on the bottom of a non-adhesive Petri dish (35- or 60-mm) and pipette 10-30 islets into the droplet.
   3. In case of different groups of islets (e.g. wild-type/knock-out), arrange multiple wells and multiple droplets so the loading can be done simultaneously.
   4. Incubate the islets in the dye working solution at room temperature in the darkness for 70-90 minutes. Do not over-incubate.
   5. Check the loading under the fluorescence microscope; the islets should gain mild fluorescence, with some cells being brighter than the rest. Rounding up of cells and nuclear localization of the dye are signs of overloading.
   6. Transfer the islets into dye-free imaging solution containing 6 mM glucose. The islets can be used for imaging immediately, but optionally the dye can be left to de-esterify for another 10-15 minutes. Islets will retain the dye for several hours and therefore can be used for imaging in several shifts.
5. Alternatively, infect the islets with the recombinant vector.
   1. Plate the islets in droplets in RPMI culturing medium (step 1.1.1) (e.g., 30 μL), to minimize the volume of vector needed.
   2. Add the vector at a ratio of approximately 10⁵ infectious units/islet that should ideally result in multiplicity of infection >2. Ideally the ratio should be optimized to the minimal ratio that would provide expression in the peripheral layer. Pre-titration (step 2.2.3) may help.
   3. Introduce 20-50 islets into the droplet and culture for 8-48 hours. (Ideally, overnight). Islets should develop a faint green fluorescence in most of the cells, without changes in cell morphology.
      NOTE: The success of the infection and expression depends on the time of exposure to the virus solution. Ideally, the virus should stay in the solution overnight but can be optionally removed after as little as 15 minutes. However, infectivity, and therefore expression, is likely to be dramatically lower.

3. Imaging Ca²⁺ dynamics

1. Immobilize the islets under the (inverted) microscope.
   1. Assemble the imaging chamber for inverted microscopy. Position the glass coverslip (thickness 1 or 1.5) inside the chamber and make sure that the glass-chamber interface is water-tight. Check that the coverslip is within the reach of the microscope objective (critical for the case of a bulky high-NA objective).
   2. Prepare the immobilization accessories. Cut small rectangles (20 mm x 20 mm) from the fine mesh and the coarse mesh. Introduce two spacer "walls" on the fine mesh using a 45-50 μm thick sticky tape. Use double layers of the spacer if the size of the islets to-be-imaged substantially exceeds the conventional 100 μm.
   3. Immerse the meshes and the weight into the imaging solution using a 35 mm Petri dish. Make sure that the plastic and the metal are wet.
   4. Under a dissection microscope, turn the fine mesh with the spacer "walls" upside down, the spacers facing upwards. Pick several islets loaded with the trappable dye or expressing the recombinant sensor with a P20 pipette and gently position them on top of the fine mesh, between the two spacers. Ensure that the mesh and the washers do not contain excessive amounts of the imaging solution on them.
   5. Pick up the mesh with the islets, using watchmaker's forceps, and position it upside down inside the imaging chamber, so that the spacers face downwards and sit directly on the chamber coverslip. Ensure that the islets are trapped between the spacers and the mesh, in the middle of the coverslip.
   6. Position the coarse mesh and the weight on top of the fine mesh within the chamber. Introduce the imaging solution into the chamber. Ensure that the islets are immobilized and ready to be imaged. Avoid excessive shaking of the chamber (small perturbations like carrying the chamber to the microscope and inserting into a heated stage are acceptable).
      NOTE: A similar immobilization arrangement can be applied for an upright system.
2. Set up the microscope.
   1. Choose the imaging mode and the objective, position the chamber with the islets from step 3.1 on the temperature-controlled stage of the microscope.
      1. Set the temperature control (ideally, between 30 °C and 36 °C) and the perifusion. For an inverted system, position the inflow lower than the outflow within the chamber, and set the outflow flux to be greater than that of the inflow (which is typically achieved by using a tubing of a wider inner diameter on a peristaltic pump).
      2. Ensure that the outflow has a minimal contact surface with the solution, so that it removes the solution in multiple sequential small droplets, avoiding long intervals of continuous solution removal. The latter is the major source of artefact in time-lapse imaging of the periodical signals as they appear as regular periodic intensity oscillations of every imaged pixel and are frequently interpreted as "slow waves".
      3. Initiate the perifusion with the imaging solution containing 3 mM glucose.
   2. Choose the light path and filters for imaging of the green fluorophores; excitation between 470 and 500 and emission between 505 and 550 would work for each of them.
   3. Run live imaging to set up the imaging parameters. Adjust the view to capture the islets of interest.
   4. Optimize the signal-to-noise ratio of the image. To that end, adjust the excitation light intensity, the exposure time and the binning. Ensure that the settings allow a distinct visualization of each cell within the islet at the minimal possible light intensity and exposure.
   5. Perform image acquisition. Depending on the task, images can be taken at 0.1 to 5 Hz. This is well below the Nyquist criteria for the fast Na⁺-driven oscillations in α- and δ-cells (>300 Hz), which means that the data is undersampled by default. However, increasing the acquisition frequency to match this demand is not feasible in multicellular/multi-islet imaging with a large field of view. GCaMP can be imaged faster, whereas Fluo4 will inevitable bleach under fast acquisition conditions.
      NOTE: Given that [Ca²⁺]_cyt oscillations in the islet cells are driven by electrical activity, using low acquisition rates may sound counterproductive. In reality, however, acquisition rates at around or above 1 Hz may be sufficient for resolving β-cell spiking behavior¹³, whereas the threshold for detection of sodium channel-driven oscillations in α- and δ-cells is well above 300 Hz. Whether the α- or δ-cell [Ca²⁺]_cyt oscillations are acquired at 1 Hz or 0.1 Hz, they will be severely undersampled and reflect Ca²⁺ handling by the cell rather than electrical activity.
      1. Check the quality of the acquired data: at 3 mM glucose, α-cell activity should be clearly visible/detectable. Ensure this is the case and proceed to full-scale time-lapse imaging.
3. Time-lapse imaging
   1. Make use of an online chart of the signal dynamics, implemented in the acquisition software if this is available. If online charting is not an option, apply a look-up-table that displays the signal intensity in the most comprehensive way (such as "rainbow").
   2. Apply the stimuli in a reversible manner: record the recovery of the signal to the basal level. Ignore the artefacts at the very start and the end of the recording; the latter may look like irreversible "increase"/"decrease" of the probe fluorescence due to changes in pH or cell death.
   3. Differentiate α-cells by oscillatory Ca²⁺ dynamics, at low glucose. Introduce adrenaline or glutamate into the bath solution, reversibly for 2-5 min. A rapid jump in [Ca²⁺]_cyt followed by a slow-down or cancellation of the oscillations will follow.
      NOTE: Adrenaline is a recognized marker compound for α-cells, that has a selective positive effect on this subpopulation of islet cells, mediated by the release of Ca²⁺ from intracellular depots¹⁸. Glutamate has been put forth as another α-cell specific agonist²³.
   4. Add/remove ghrelin, which has been recently reported to activate δ-cell selectively^19,20. Observe a rapid reversible increase in [Ca²⁺]_cyt in a small subpopulation of islet cells.
   5. Add/remove 20 mM glucose. Observe a coordinated oscillatory response in the β-cell subpopulation. Note the response of cells that have earlier been activated by adrenaline or glutamate and ghrelin.
   6. Save the image sequence. Consider using "AutoSave" during the recording.

4. Analyzing the data

1. Analyze the time-lapse image.
   1. Import the time-lapse image into an image analysis software, such as an open-source ImageJ/FIJI.
   2. If substantial/rapid movement has occurred during the recording, discard the data as unrepairable. Use StackReg or TurboReg plugins²⁴ to correct minor drifts.
   3. Create a mask image for cell detection and region of interest (ROI) mapping. The preferred way to achieve this would be to make a stack image using one of the functions such as "average intensity" or "maximal intensity". The function to be used is the one that will provide the best recognition of individual cells.
   4. Threshold the mask image, remove all the data outside the islet region. The function works in automatic mode for 32-bit images.
   5. Detect maxima in the threshold image. The maxima can be represented by points, regions or even sectors, if the image is dense.
   6. Apply the 'find maxima" function without any size restrictions and paste the detected maxima into the region of interest (ROI) editor.
   7. Smooth/interpolate each of the ROIs mapped; possibly, the ROIs will require expansion. A simple script can be written for (4.1.6-4.1.8) and run several times to provide best cell detection results. Several ROIs may overlap but this is rare.
   8. Analyze the position of the ROIs and paste the respective X and Y data into the electronic table software (e.g., Microsoft Excel).
   9. Analyze the grey intensity vs. time for all the ROIs and paste the data into the electronic table software.
2. Analyze the numerical data.
   1. Import the data into the data analysis software. Depending on the choice of the software and the duration/sampling/size of the experiment, this can be a simple copy-paste operation or a standalone procedure. Ensure the arrangement and storage of the numerical data.
   2. Import the time-stamps or the time notes, if available.
   3. Normalize the raw fluorescence intensity data ("F") to the initial value of fluorescence ("F₀"). This does not need to be the fluorescence in the very first point but could be the average of several first points. The normalization should reduce the variability of the data and, in an ideal case (no drifting baseline) result in an analyzable dataset ("F/F₀").
   4. If the cell-to-cell variability of the F/F₀ dataset is still substantial (long recording, bleaching), perform the baseline correction. To that end, define a 'control' region, i.e. the range of time during which the control solution (for mouse pancreatic islets, 3 mM glucose without any agonists/antagonists) was applied.
      1. If the control region has a clear non-oscillatory signal, assume that F/F₀ was returning to the initial value (F/F₀=1) after each application of the control solution. Correct the time-lapse data for each cell by splitting the data into segments, separated by the points when the control solution was added, and applying linear correction to each segment. Do not use polynomial or other nonlinear correction as this results in artefacts.
      2. If the control range has clear oscillations or additional factors (such as FAD autofluorescence) are present, use a spike detection algorithm¹⁷. A trivial and fast run-around for that is a maxima-sensitive wavelet transform (Figure 3A).
   5. Quantify the data. Although Ca²⁺ is a highly dynamic signal, presenting the Ca²⁺ data in terms of absolute F/F₀ values is widely acceptable in biomedical literature. If the results from multiple experiments need to be compared, choose a metric.
      1. Measure the frequency of Ca²⁺ spikes (Figure 4A,D), and its response to addition of (ant)agonists. To that end, split the recording into equal time intervals and compute the timecourse of the partial frequency (spiking frequency in each of the intervals) by counting spikes within the interval and normalizing to the interval duration.
      2. Alternatively, set the threshold and compute the plateau fraction (pf) for each of the intervals defined above (Figure 4B,F). The fraction indicates the percentage of time within the interval that the cell spent in the "excited" state.
      3. Alternatively, compute the partial area under the curve (pAUC) for each of the intervals defined above (Figure 4C,G). This metric is sensitive to changes in both frequency and amplitude of spiking.
         NOTE: One caveat for measuring frequency is its lack of sensitivity for the spike duration and poor stability. As the data is heavily undersampled vs. the electrical spiking, the number of spikes per interval is quite small and hence a sole extra spike can dramatically affect the result. The 'bottleneck' of the pAUC is its sensitivity to changes in the baseline. Although less prone to artefacts and more sensitive to changes in [Ca²⁺]_cyt than frequency, pAUC nevertheless is not very informative about the nature of Ca²⁺ dynamics. Plateau fraction is an expansion of the open probability concept to the whole-cell system. It is less robust than pAUC though, due to its dependence on the threshold value.

Representative Results

Islets load fairly well with the trappable dyes (Figure 1A), unless the lipid composition of the membrane has been affected (e.g., by chronic exposure to fatty acids). The human adenovirus type 5 (Ad5) vector also targets all islet cells (Figure 1B). Problems may arise when more than one recombinant sensor is being expressed in the same cell. Furthermore, islets are typically very well immobilized using the technology described above, which provides exceptional stability and solution access.

Ca²⁺ spiking in α-cells can be readily detected at low glucose levels (Figure 2). There is a high cell-by-cell correlation between the activity at low glucose and the response to adrenaline and glutamate. Ghrelin activates some adrenaline-responsive cells (α-cells?) at low glucose yet it has no effect on Ca²⁺ dynamics in most of the cells that are activated by low glucose (β-cells).

When analyzed in terms of partial frequency (Figure 4A,C), adrenaline- or ghrelin-stimulated cells display a substantial increase under the all-or-nothing conditions. That is, a cell with low basal activity that gets activated by adrenaline or ghrelin exhibits a dramatic increase in this metric. However, the overall changes between basal spiking and the adrenaline effect are very subtle (Figure 4A,C). In contrast, partial AUC is sensitive to the changes introduced by adrenaline in all cells, even when basal activity is high (Figure 4B,D).

Figure 1: Loading of the trappable dye and expression of the recombinant sensor in islets. Typical mouse islets loaded with the trappable dye Fluo-4 (A) or expressing the recombinant sensor GCaMP6 in the peripheral layer of cells (B) or in the deeper layer (C). Polar tracer sulforhodamine B (SRB, shown as white) has been utilized to outline individual cells within each islet²⁵. (D) Representative kinetics of Ca²⁺ in response to glucose recorded from individual cells within the islet using Fluo4. Note the heterogeneity within minor cell populations. Please click here to view a larger version of this figure.

Figure 2: Typical Ca²⁺ response of islet cells to various stimuli. Typical α-cell (A) and δ-cell (B) Ca²⁺ dynamics, in response to adrenaline, glutamate, ghrelin, glucose. (C)-(D) Heat maps of the islet cell response showing adrenaline-positive (C) and ghrelin-positive (D) subpopulations. Please click here to view a larger version of this figure.

Figure 3: Baseline correction. Please click here to view a larger version of this figure.

Figure 4: Analysis of the time-lapse data. Analysis of the Ca²⁺ dynamics in α-cells. Partial frequency (A), plateau fraction (B) and area under the curve (C) of an α-cells [Ca²⁺]_i trace. Populational [Ca²⁺]_i data from a mouse pancreatic islet expressed as raw (F/F₀) (D), partial frequency (E), plateau fraction (F) and area under the curve (G). Please click here to view a larger version of this figure.

Discussion

There are three stages in the protocol that are critical for overall success. Successful injection of the Liberase enzyme into the bile duct determines not just the quantitative success of the isolation procedure but also impacts the quality of the isolated islets. Uninflated pancreata may result in lack of some important metabolic responses in the isolated islets. Secondly, the loading of the dye/the expression of the sensor defines the signal-to-noise ratio of the time-lapse recording. Signals are absent or attenuated in overloaded islets. Lastly, successful and dense positioning of the tissue inside the imaging chamber is a defining moment for meaningful and analyzable experiments. Poorly positioned or moving tissue results in wasted experimental time and/or unclear data.

The method can be modified to account for multiple signals (using the confocal system) and multiple groups of islets (e.g., of different genotype). Imaging of multiple signals assumes delivery of a second sensor into each cell of the islet, spectrally compatible with the Ca²⁺ reporter (such as a pH sensor SNARF5f^26,27). To that end, islets can be co-loaded/co-infected with Ca²⁺ and pH sensors, which are then sequentially imaged within each time-frame.

Imaging the signal in groups of islets with single-cell resolution requires using a wide field-of-view objective. The objective is likely to have lower magnification and numerical aperture (NA), thereby reducing the spatial resolution. Due to the increased depth-of-focus of the low-NA objective, imaging can be performed on a wide-field system. The disadvantages of this arrangement are cell-cell cross-contamination of the light signal and reduced ability to image the 3D signal (e.g., mice expressing the Ca²⁺ sensor under the insulin promoters). At the same time, signal expressed from the surface islet cells can be perfectly resolved with high temporal resolution from groups including tens to hundreds of islets¹⁸.

Although it may sound unpleasant but performing image analysis and numerical data analysis in separate software packages is a good idea. At the current time, ImageJ/FIJI dominates scientific image analysis. The most popular environments for scientific coding are Python and Matlab, yet there are also known efforts to analyze the Ca²⁺ data in R²⁸. Best usability is provided by more niche packages like IgorPro. Our choice is to prototype in Matlab/Python and then implement the code in IgorPro for 'pipeline' use. Adapting signal analysis packages for electrophysiology (e.g., Clampfit, Neuroexplorer) for analytical needs can be useful for single-cell imaging but is difficult to scale up. Many options provided by such packages are not applicable for islet imaging because of low sampling rate.

It is important to remember that this methodology is limited by a number of factors. Firstly, as mentioned above, imaging is largely based on undersampling the data, meaning it does not indicate and therefore cannot be directly compared to electrical activity of the cell. Secondly, the data comes from the islet periphery and does not reflect important coupling processes that are, generally speaking, three-dimensional. Thirdly, the level of loading/expression affects the perception of the sensor intensity. Lastly, activation of less well-researched islet cell subpopulations (e.g., PP cells & ϵ-cells) by the marker compounds cannot be ruled out, although due to the low numbers of these cells within the islet any potential contamination will be minimal.

The method is a true 'champion' in terms of visual effect, as the oscillatory processes deliver a strong impression of a genuinely living tissue. Applied to minor cell subpopulations, the method probes the function of each one reliably, allowing identification of subgroups and reflecting the heterogeneity.

Calcium dynamics has been studied in pancreatic islet β-cells for over 40 years, mostly driven by the progress in acquisition/detection technology. Early studies used atomic absorption spectroscopy²⁹, but it was not until the arrival of fluorescent Ca²⁺ sensors³⁰ that detailed kinetics could be resolved in individual islet cells, using photometry^31,32,33. Soon after that, the spatial component of Ca²⁺ kinetics was improved as Ca²⁺ imaging^34,35,36 became a routine technology, thanks to the then newly available charge-coupled device (CCD) detectors. The problem of out-of-focus light, that hampered imaging the signal from individual cells within the tissue, was then resolved in the mid-1990s via laser scanning confocal microscopy (LSCM)³⁷ and total internal reflection fluorescence microscopy (TIRFM)³⁸. Both methods, complemented by the arrival of a new generation of fluorescent Ca²⁺ sensors excitable with a 488 nm laser, have been successfully used to image Ca²⁺ dynamics in the islet cell subpopulations^39,40,41.

The new century brought forward two new trends that stemmed from neuroscience-related technological developments. Firstly, recombinant fluorescent sensors based on circular permutation of GFP variants substantially increased signal-to-noise ratio for Ca²⁺ detection, effectively bringing the studies to the level of large cell populations, in which the dynamics of [Ca²⁺]_cyt in every cell could be resolved. Secondly, use of tissue-specific promoters allowed targeting sensor expression to minor subpopulations.

Although generally thought to reflect the developments in neuroscience, studies on islet Ca²⁺ dynamics have two key differences. Firstly, technologically, any in vivo imaging of islet signaling is more complex than imaging in the brain due to the unpredictable anatomy of the pancreas and the islets' location⁴². Secondly, excellent electrical coupling between the islet β-cells essentially renders islets into electrically inert populations displaying a seemingly perfect all-or-nothing response to the high glucose stimulus. We believe that studies of [Ca²⁺]_i kinetics in minor islet subpopulations, such as δ-cells, based on tissue-specific targeting are likely to broaden our knowledge of their pharmacology/physiology. At the same time, highly sensitive probes allow expanding the statistical power of such measurements, accounting to islet-to-islet variability and allowing imaging of islets from different groups within one parallel experiment.

Materials

Name	Company	Catalog Number	Comments
40x/1.3 objective
Axiovert 200 microscope
emission
Excitation
Fetal bovine serum	Sigma-Aldrich	F7524-500ML
Fluo4	Thermo Fisher (Life Technologies)	F14201
GCaMP6f, in (human type 5) adenoviral vector	Vector Biolabs	1910
Hanks' solution	Thermo Fisher (GibCo, Life Technologies)
Liberase	Sigma-Aldrich	5401020001
penicillin/streptomycin	Thermo Fisher (GibCo, Life Technologies)	15140122
RPMI medium	Thermo Fisher (GibCo, Life Technologies)	61870044
Zeiss LSM510-META confocal system	Carl Zeiss