Glucose and NAADP trigger elementary intracellular β-cell Ca2+ signals

Pancreatic β-cells release insulin upon a rise in blood glucose. The precise mechanisms of stimulus-secretion coupling, and its failure in Diabetes Mellitus Type 2, remain to be elucidated. The consensus model, as well as a class of currently prescribed anti-diabetic drugs, are based around the observation that glucose-evoked ATP production in β-cells leads to closure of cell membrane ATP-gated potassium (KATP) channels, plasma membrane depolarisation, Ca2+ influx, and finally the exocytosis of insulin granules. However, it has been demonstrated by the inactivation of this pathway using genetic and pharmacological means that closure of the KATP channel alone may not be sufficient to explain all β-cell responses to glucose elevation. We have previously proposed that NAADP-evoked Ca2+ release is an important step in stimulus-secretion coupling in pancreatic β-cells. Here we show using total internal reflection fluorescence (TIRF) microscopy that glucose as well as the Ca2+ mobilising messenger nicotinic acid adenine dinucleotide phosphate (NAADP), known to operate in β-cells, lead to highly localised elementary intracellular Ca2+ signals. These were found to be obscured by measurements of global Ca2+ signals and the action of powerful SERCA-based sequestration mechanisms at the endoplasmic reticulum (ER). Building on our previous work demonstrating that NAADP-evoked Ca2+ release is an important step in stimulus-secretion coupling in pancreatic β-cells, we provide here the first demonstration of elementary Ca2+ signals in response to NAADP, whose occurrence was previously suspected. Optical quantal analysis of these events reveals a unitary event amplitude equivalent to that of known elementary Ca2+ signalling events, inositol trisphosphate (IP3) receptor mediated blips, and ryanodine receptor mediated quarks. We propose that a mechanism based on these highly localised intracellular Ca2+ signalling events mediated by NAADP may initially operate in β-cells when they respond to elevations in blood glucose.

were avoided, might reveal their substructure, akin to the sparks and puffs observed for IP 3 and ryanodine receptors, respectively 16,17,18 . A feature of Ca 2+ signalling dynamics in pancreatic β-cells is their heterogeneity upon stimulation, with the recent proposal that some cells in the islet act as Ca 2+ signalling hubs or pacemakers 19 , whilst others follow through gap-junctional or paracrine signalling mechanisms. However, heterogeneity is also seen in isolated cells. Indeed, Ca 2+ responses to glucose in β-cells have been described as having a typical triphasic shape often seen in parallel with, and assumed to be the result of, simultaneously measured changes in membrane potential, which have a similar pattern and are in synchrony with the cytosolic Ca 2+ increase 15 : Phase 0 consists of a 'dip' , or initial decrease, in cytosolic free Ca 2+ resulting from increased sarco-/endoplasmic reticulum Ca 2+ ATPase (SERCA) activity transporting Ca 2+ into the endoplasmic reticulum (ER) in response to the rising ATP concentration following glucose metabolism. Phase 1 constitutes a transient rise in Ca 2+ associated with L-type Ca 2+ channel activation and Ca 2+ -induced Ca 2+ release (CICR) from intracellular stores. Phase 2 are Ca 2+ oscillations superimposed on a steadily elevated plateau thought to be the result of Ca 2+ influx through L-type channels. While this standardised model is useful; it has been demonstrated that β-cells display more complex responses to glucose and other stimuli such as GLP-1, insulin, and nutrients including amino acids 20,21 . A classification of primary human β-cell autocrine Ca 2+ responses to insulin shows that β-cells display a variety of equally common but different Ca 2+ signals and no clear 'standard' response 22 .

Results
We report here sub-membrane Ca 2+ transients evoked by 16.5 mM glucose in mouse primary pancreatic β-cells, measured using evanescent-wave TIRF microscopy. Recordings were made from 1,017 cells. Figure 1 shows the classification of typical responses of 239 cells from male WT mice of one genetic background, to avoid additional confounding factors. There was a clear heterogeneity of responses. While over 96% of cells exhibited a clear elevation of sub-membrane Ca 2+ (calculated as ∆F/F 0 , where ∆F is the change in fluorescence intensity from pre-stimulation F 0 ), ∼60% resembled the standardised triphasic profile, with the remaining responding cells described by 3 further classifications (cf. Fig. 1). These data are consistent with the notion that there are β-cells with different patterns of expression of Ca 2+ channels, which may serve multiple functions within the islet 19 . Treatment of mouse primary pancreatic β-cells with extracellular NAADP-AM (10 µM) resulted in similar Ca 2+ responses to those observed with glucose described above. Ca 2+ oscillations could be resolved that culminated in a raised plateau of elevated [Ca 2+ ] i in 89% of cells (see, for example, a 'type 4' response elicited by NAADP-AM in Fig. S1). Response type distributions for stimulation with NAADP-AM were as follows: Type 1, 27%; Type 2, 0%; Type 3, 5%; Type 4, 59%, no response, 9%, n = 53.
If Ca 2+ release from sub-membrane stores triggered the global Ca 2+ response, an increase in Ca 2+ could be expected to occur first in the vicinity of the cell membrane, the location of the cell's secretory vesicles which form part of the acidic organelle continuum in the β-cell 23 , before recruiting a global Ca 2+ response 24 . Lysotracker was utilised to confirm the presence of acidic stores within the TIRF plane (see Fig. S2). In parallel recordings using TIRF and standard epifluorescence (to monitor global Ca 2+ ), however, the two transients were largely superimposable (see supplementary data, Fig. S3). Thus if there were sub-membrane Ca 2+ release events preceding the global Ca 2+ response, they were either very rapid and too small to be detected by the current protocol, masked by concomitant larger L-type Ca 2+ channel-mediated influx, or obscured by increased SERCA pump activity following enhanced glucose metabolism to ATP.
In order to minimise these possible confounding factors, cells were pre-incubated in recording medium containing 1.7 mM Ca 2+ , 3 mM glucose, and 1 μM of the irreversible SERCA inhibitor, thapsigargin. Immediately before recording, the extracellular fluid was exchanged for Ca 2+ -free medium, containing thapsigargin and either 100 µM or 5 mM EGTA. The removal of Ca 2+ immediately before recording prevented non-ER stores, which may rely on Ca 2+ influx and ER Ca 2+ transfer for filling 25,26 , from run-down during the incubation period. Cells were then challenged with either 100 nM NAADP-AM or 16.5 mM extracellular glucose. Over the ∼10 min period following either challenge, cells showed clear very brief sub-membrane Ca 2+ transients with maximum amplitudes > 2∆F/F 0 . These show comparable kinetics to those of Ca 2+ puffs evoked by IP 3 27 . The movie (supplementary information, Movie 1) illustrates recordings of these events, and examples are shown in Fig. 2a-i. Their localised nature is illustrated in the maximum ∆F/F 0 trace (Fig. 2m) and the 3D fluorescence intensity plots shown in Fig. 2 j-l. Similar localized Ca 2+ signals were observed in extracellular solutions containing 100 µM and 5 mM EGTA (cf Fig. S4), strongly suggesting that Ca 2+ release from intracellular stores but not Ca 2+ influx are involved in the generation of these elementary Ca 2+ events. These local transients were also recorded at a higher acquisition rate (46 Hz) (Fig. S5, Movie 2), giving additional resolution of the events captured at our standard experimental recording rates of 3.3 Hz.
To resolve these individual events in more detail, a spark detection algorithm was used (see Methods) which also removed the underlying 'ramp' in cell Ca 2+ that can be seen in Fig. 2m. There is an increasing frequency of events following application of 100 nM NAADP-AM or 16.5 mM glucose to a maximum; followed by a gradual decline (Figs. 3a, b, S6). This is in accordance with the self-limiting nature of these types of signals when stores cannot be replenished during sustained stimulation 28 or desensitization to signals such as NAADP 22,29 . To quantify the changes in frequency, the mean percentage of frames showing at least one spike (defined as an event 2 standard deviations above mean baseline ∆F/F 0 ) was determined for the baseline (last 80-100 frames before stimulus) and after the addition of raised glucose or NAADP-AM. As illustrated in Fig. 3e, f, both 100 nM NAADP-AM and 16.5 mM glucose evoked an approximately ten-fold increase in spike frequency.
Having unmasked discrete Ca 2+ signals in response to a high concentration of glucose (16.5 mM) after thapsigargin pre-incubation, this raises the question of whether similar responses could be evoked by a smaller increase in extracellular glucose without the use of thapsigargin. We examined whether a doubling of glucose from 3 to www.nature.com/scientificreports/ 6 mM (a stimulatory concentration within the physiological range) without thapsigargin pre-incubation, would still allow us to detect visible Ca 2+ events with TIRF. We reasoned that at lower concentrations of glucose, less ATP would be produced, resulting in lower SERCA activity, which would otherwise obscure detection of local Ca 2+ events. As shown in Fig. 3c, the effects of stimulation with low glucose in the absence of thapsigargin resemble those of stimulation with high glucose in the presence of thapsigargin. The percentage of frames containing spikes is significantly higher after raising extracellular glucose to 6 mM (15%) than at baseline (4%; Fig. 3g).
Using epifluorescence microscopy, no local Ca 2+ release events could be detected (Fig. 3d).
To analyse the nature of the elementary Ca 2+ events, optical quantal analysis was carried out on images like those shown in Fig. 4d, e (Fig. 4a, b show the same cells under brightfield microscopy and TIRF, respectively). Plotting the frame-maxima across all cells stimulated with NAADP-AM results in the frequency distribution depicted in Fig. 4f. Modes are visible in the 'tail' of this decaying function which occur at a period of around 0.05 ∆F/F 0 (see figure inset). This order of magnitude is comparable to optically-assessed Ca 2+ blips through IP 3 receptors (0.1 ∆F/F 0 27 ) and the smallest imaged IP 3 -evoked Ca 2+ signals in pancreatic acinar cells (< 0.1∆F/F 0 30 ). To demonstrate the quantal nature of the Ca 2+ responses at the level of the individual cell, responses from a representative cell cluster are shown in Fig. 4g. Specific regional areas of interest (ROIs) within the cell were chosen by hand on the basis of their containing at least one Ca 2+ signalling event. These ROIs were then analysed for their mean amplitudes after subtraction of baseline fluorescence. Events were included if they lay above baseline mean + 2 SDs) of that particular cell. The frequency distribution displays modal behaviour as illustrated by the poly-Gaussian function fitted to it, and the putative unitary event amplitude for this cell cluster is 0.03 ∆F/F 0 .

Discussion
The above experiments demonstrate that β-cells show great variability in their Ca 2+ responses and a high level of spontaneous activity; supporting the notion of a β-cell's 'Ca 2+ fingerprint' 31 . As β-cells are frequently examined in artificially amplified resting and active states during experiments 6,32 , we have shown here that stimulation with the high glucose concentrations often employed, leading to high SERCA activity 33 , obscures more subtle Ca 2+ changes taking place within the cell. It is probable that multiple stores and channels play a part in this process 32,34 . The fact that the events occur after pre-incubation with thapsigargin, ruling out the ER as a source, are most striking when triggered with NAADP-AM which targets acidic stores, and are localised just beneath the membrane (the primary location of insulin granules, a subset of the acidic organelles in the β-cell 35 ), is a strong suggestion that the source of these events are acidic Ca 2+ storage organelles. However, further pharmacological studies are required to fully delineate the origin of these signals.
With the discovery of two-pore channels (TPCs) as the putative target for NAADP 36,37 , and the present discovery of elementary Ca 2+ release events in response to NAADP; it is likely that NAADP-evoked Ca 2+ signals from acidic stores are built from elementary events via activation of TPCs. It was suggested in 1996 that NAADPmediated Ca 2+ release was quantal on the basis of its graded release from sea urchin homogenate 38 and in intact sea urchin eggs localized Ca 2+ responses ascribed to the osmotic lysis of acidic stores by GPN 39 ; 25 years on, we may have evidence in a mammalian cell type that a similar principle operates. Thus with the existing sets of  In addition to demonstrating elementary Ca 2+ signals in response to NAADP for the first time, the present study also suggests a potential role for these events in stimulus-secretion coupling in β-cells. NAADP has been shown to elicit Ca 2+ signals and insulin release in mouse pancreatic β-cells 7,22,10,11 . Whilst there is argument over the identity of NAADP's target in the β-cells 7,40,41 , mutations in the two-pore channel gene (TPCN2), the potential principal target for NAADP, have been implied in the inheritance of diabetes type 2 in humans 42 . Localized Ca 2+ signals from acidic stores have been proposed to cause depolarisation by activating calcium-dependent cation channels in the plasma membrane as we have previously observed 7,43 , such as TRPM4 7 and TRPM5 44,46,47 . NAADP applied through a patch-pipette evoked small oscillatory cation currents which were preceded by small Ca 2+ transients 7 which were abolished in cells from Tpcn2 −/− mice. Importantly, elevating glucose concentrations www.nature.com/scientificreports/ evoked similar cation currents, which along with those evoked by NAADP were inhibited by the NAADP antagonist, Ned-19 7 . We have now imaged these localized Ca 2+ events with our TIRF methodologies in the current study. However, local Ca 2+ release from acidic stores is also known to trigger CICR in many cells, likely involving membrane contact sites with the ER 48 . In yet other cell types, NAADP was shown to induce localized Ca 2+ release from secretory granules to initiate their own exocytosis 49 . In pancreatic β-cells, ER Ca 2+ leak and subsequent uptake into mitochondria via the mitochondrial Ca 2+ uniporter (MCU) complex have been proposed to prime www.nature.com/scientificreports/ ATP synthesis 50 . Since it has been suggested that TPCs can be blocked by ATP 51 , there may be a complex interplay between Ca 2+ release from acidic stores and the dynamics of ATP concentrations at the subcellular level. Thus we propose that during stimulus-secretion coupling in β-cells K ATP channel closure induced by a rise in local ATP levels increases membrane resistance allowing small cation currents, activated by localized Ca 2+ signals (a summation of elementary Ca 2+ events from intracellular stores and CICR) to initiate depolarization of the plasma membrane. This in turn results in the opening of L-type Ca 2+ channels whose mediation of larger globalized Ca 2+ signals triggers insulin granule exocytosis. This model differs from, but contains elements of, each of the three models of a K ATP channel independent pathway previously discussed. The present investigation suggests that glucose initially generates localized Ca 2+ signalling events based on Ca 2+ release from non-ER, likely acidic, stores prior to the influx of Ca 2+ through VGCCs. We propose that some of the elementary trigger events involved are likely to be mediated by NAADP as we have previously suggested 7,24,45 .

Methods
Primary β-cell culture. Mouse pancreatic islets from 10-14 week-old WT male mice (of a mixed C57BL/6;129P2 background obtained from the European Mouse Mutant Archive, EMMA) were isolated as described previously 52 . Islets were dispersed into single cells or cell clusters and plated onto poly-L-lysine coated glass coverslips (Menzel) and incubated in cellstar dishes (Greiner) at 37 °C for 4-7 h before adding cell culture medium (RPMI 1640, -glucose + glutamine, Gibco), supplemented with penstrep (10,000 U/ml penicillin/10,000 μg/ml streptomycin, Gibco) and 10% fetal bovine serum (FBS, Gibco) containing 10 mM glucose and incubated for a further 15-17 h before first use. Cells were loaded with 500 nM fluo 4-AM (Invitrogen) for 1 h in the dark, before being washed with imaging buffer (NaCl 130 mM, KCl 5.2 mM, MgCl 2 1 mM, HEPES 10 mM, CaCl 2 1.7 mM; pH: 7.4, 280-340 mOsm/kg) containing 3 mM glucose and left in this for 10 min before start of recording. Experiments were conducted exclusively at 37 °C, and with a baseline glucose concentration of 3 mM, simulating physiological conditions. Experiments to demonstrate localised elementary events were conducted on both single cells and small clusters; at an estimated ratio of 50% cells, 50% clusters. Clusters lend themselves to TIRF experiments, as they provide a large area positioned in the same focal-plane. Dispersed cells are usually at marginally different focal planes due to the inherent curvature of the glass coverslip. In studies using Ca 2+ -free medium, extracellular Ca 2+ was re-admitted at the end of the experiment to visualise a global response as a verification of the cell's viability. In experiments to visualise acidic stores, cells were preincubated for 10-30 min with LysoTracker Red (Invitrogen) at 200 nM. NAADP-AM was synthesised as previously described 13 . Batches varied in potency due to different efficiencies of synthesis and storage degradation as determined by HPLC analysis.
Imaging. Cells were excited with an Argon-Ion laser (Andor DU-897, 40 mW; Melles Griot) at 488 nm, and images were obtained using a Nikon Evanescent Wave Imaging System; an Inverted Total Internal Reflection Microscope (Nikon Eclipse Ti) equipped with 60 × and 100 × CFI Apochromat TIRF Series oil-immersion lenses. These lenses have a numerical aperture of 1.49, which allowed for a maximal incident angle of 76.87° calculated by α ≥ sin −1 (n 2 /n 1 ) , n 1 > n 2 , where n 1 is the refractive index of the cover glass (1.53), and n 2 the numerical aperture of the lens (1.49). The exact angle for experimentation was determined using a Bertrand lens and the beam adjusted. This allowed for an evanescent wave illuminating the specimen to a depth of around 100 nm. All parameters were controlled using NIS Elements AR 4.0 (Nikon). Images were acquired at a rate of 3.3 Hz for single channel recordings (i.e. only TIRF or Epifluorescence) and at a pre-programmed rate for dual channel recordings (Frame Rate Epi: 1 Hz. Frame Rate TIRF: 3.3 Hz). Data in Fig. S5 and Movie 2 were acquired using NIS Elements RAM capture mode (acquisition rate ~ 46 Hz). A CCD camera (Andor iXon +) was used to capture emitted fluorescence at 515-555 nm (Binning: 1 × 1, Exposure: 300 ms, Multiplier: 89, Readout Speed: 10 Hz, Conversion Gain: 1x, Dimensions: 512 × 512, 160 nm/pixel).
Analysis. For quantal analysis, fluorescence change across each whole cell (selected as an area of interest, ROI, see Fig. 4c) was analysed by first normalising each image (time frame) pixel by pixel with respect to the average fluorescence across the last 80-100 frames before cell stimulation with glucose or NAADP-AM (cells stimulated using glucose were not used for quantal analysis, as it is an unspecific stimulus and may trigger events via multiple pathways). To eliminate the moving baseline (ramp), a frame by frame subtraction function was used. Of these normalised, baseline-controlled images, the maximum intensity within each cell was measured. A similar algorithm has been used for 2D images previously 53 , and while this paper was in preparation, an automated system applying a sophisticated version of it was released 54 . From the measured maximum intensities for each cell over time, frequency histograms were compiled to depict the quantal nature of the events, where events more than 2 standard deviations above the baseline mean were considered genuine. This is a standard estimate of minimum spark amplitude 30 . All analysis was conducted in NIS Elements AR 4.0 (Nikon) and MS Excel 14 (Microsoft). Figures were prepared in Illustrator CS6 (Adobe). Ca 2+ traces are of fluorescence change relative to baseline mean fluorescence (∆F/F 0 ). Baseline mean fluorescence (F 0 ) was calculated from the fluorescence of the last 80-100 frames before stimulation. Traces are representative, as indicated in figure legends. Ca 2+ fluorescence images are pseudo-coloured so that changes in colour reflect changes in fluorescence. Warmer colours represent higher levels of fluorescence. Statistical analysis was conducted in MS Excel 14 and SPSS 19 (IBM). Student's t-tests (paired or unpaired, one-or two-tailed, as applicable) were used to determine the statistical significance of observed effects (p < 0.05, p < 0.01, as stated). Charts illustrating statistical differences between groups depict mean ± standard error of the mean (SEM) unless stated otherwise. Videos for supplementary information were prepared from Nikon .nd2 files and rendered using Premiere Pro CS6 (Adobe).