Male C57BL6 mice aged 4–8 weeks obtained from Harlan Laboratories (Indianapolis, IN) were used for all experiments. The animals were cared for according to the guidelines of the Vanderbilt Institutional Animal Care and Use Committee.

Islets were isolated in Hanks Balanced Salt solution, and HEPES-buffered Krebs Ringer Bicarbonate solution (KRBH) was used for the major part of the static incubations in secretion experiments. The components of KRBH are as follows: 128.8 mM NaCl; 4.8 mM KCl; 1.2 mM KH₂PO₄; 1.2 mM MgSO₄; 2.5 mM CaCl₂; 5 mM NaHCO₃^- and 10 mM HEPES. In the Cl^--free KRBH medium used for intracellular alkalinization, gluconate salts were used in place of Cl^- salts. The medium pH was maintained at 7.4, except in the media used for intracellular alkalinization where the pH was raised above 8.3 by addition of NaOH. Basal KRBH used for pre-incubation and non-stimulated controls contained 2.8 mM glucose, while the stimulating media contained either 16.7 mM glucose, or other secretagogues as indicated in the presence of 2.8 mM glucose. In pH_i-alteration experiments, 40 μM di-methyl amiloride (DMA) was added to the medium to produce intracellular acidification, and a Cl^--free medium, a high pH (>8.3) medium, or a combination of both, were used for intracellular alkalinization. In preparation for imaging experiments, islets were cultured in RPMI 1640 culture medium supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 0.1 g/L streptomycin and 11 mM glucose.

In addition to 16.7 mM glucose, the following mitochondrial secretagogues were used to stimulate direct insulin secretion. a) methyl pyruvate (MP), the methylated form of the glycolytic end product of glucose, b) leucine, an amino acid metabolized solely in the mitochondria, c) αKIC, a metabolic product of leucine, and d) 2-amino-bicyclo[2,2,1]heptane-2-carboxylic acid (BCH), a non-metabolizable analog of leucine which enhances TCA cycle activity only through allosteric activation of glutamate dehydrogenase (GDH) 363738. None of these compounds are metabolized in the cytosol, and their metabolic effects are exerted solely in the mitochondria. The non-mitochondrial secretagogues used were a) high K⁺, which directly stimulates insulin secretion through depolarization-induced Ca²⁺-influx, activating the first phase of insulin release, and b) glucose combined with inhibitors that prohibit the entry of glycolytically-derived pyruvate and NAD(P)H into mitochondria, thus eliminating the mitochondrial component of glucose metabolism while leaving the glycolytic component intact 39. The inhibitors included α-hydroxycyanocinnamic acid (CHC), an inhibitor of mitochondrial pyruvate transport, and amino-oxy acetic acid (AOA), an inhibitor of the malate aspartate shuttle that transports glycolytic NAD(P)H into mitochondria 39.

A modified version of the collagenase digestion method described by Lacy and Kostianovsky 40 was used. Mice were anesthetized with intra-peritoneal injection of Ketamine/Xylazine (80/20 mg/Kg). Pancreas was removed, placed in ice-cold Hanks solution and minced with scissors. Collagenase (3 mg/ml) was added and the mixture shaken in a 37°C water bath until the tissue was adequately digested. The mixture was then centrifuged, supernatant removed and the pellet re-suspended in Hanks solution. Centrifugation and re-suspension were repeated several times to remove exocrine tissue. The final pellet was re-suspended either in basal KRBH medium for secretion experiments or RPMI medium for islet culture. Islets were hand picked under a stereomicroscope.

The method described by Arkhammar et al. 1998 41 was used with minor modifications 3942. 35 mm culture dishes with glass-bottomed wells (Mat-Tek corporation) were used. The dishes were pre-prepared by coating the wells with human extracellular matrix (BD Biosciences). Freshly isolated islets were placed carefully in each well, covered with RPMI medium containing 11 mM glucose, and cultured at 37°C in 95% O₂ and 5% CO₂. Under these conditions, the cells in the islet spread out within 14 days, greatly reducing the islet thickness and making it particularly suitable for imaging with confocal microscopy. An advantage of this culture technique is that it does not significantly disrupt islet architecture, thus allowing islet cells to maintain normal functions. Although the islet spreads out over the matrix, islet cells do not separate from each other, and the gap junctions between cells continue to function. Islets cultured for two weeks under these conditions exhibit normal responses of Ca²⁺, NAD(P)H and insulin release to glucose stimulation 394142, providing an excellent model for imaging studies whose results can be safely extrapolated to fresh islets.

The changes in islet pH_i produced by each secretagogue and by different pH_i-alteration techniques were monitored by confocal microscopy, using carboxy-seminaphthorhodofluor-5 (SNARF5) 35, a pH-sensitive fluorescent indicator. Prior to imaging, cultured islets were maintained in RPMI medium containing 5 mM glucose for 48 hours. On the day of the experiment, RPMI medium was removed, and islets were washed and placed in basal KRBH medium. SNARF5-AM (5 μM final concentration) was added and incubated for one hour at 37°C. Loaded islets were placed on a warmed stage in a humidified and temperature-controlled chamber at 37°C, and monitored with a F-Fluar 40 × 1.3 NA oil immersion lens of a LSM510 confocal laser-scanning microscope (Zeiss). Islets were excited at 514 nm with an argon laser, and the emission fluorescence was collected in the band-widths 568–589 nm and 621–643 nm (peak emission at 580 and 630 nm) using the Meta detector (Zeiss). Time-series images (2.56 μs/pixel) were collected for 5–20 minutes as was suitable for each experiment. A stable baseline was obtained before the actual recording for each experimental condition started. LSM software was used to calculate the ratio between the two emission fluorescence values from selected well-loaded regions in each islet. This ratio was proportional to the islet pH_i. The results were analyzed using LSM software, Graphpad Prism, and Microsoft Excel. One representative recording for each experimental condition is shown in the results section, and the value n denotes the number of recordings done with different islets for each condition. The pH_i change for each condition is the difference between the average pH_i of the baseline and the average pH_i over a stable region of the recording after treatment. The pH_i changes from n recordings were averaged to obtain the "average pH_i change" shown in the results section and in the legend for each figure. A standard curve was prepared by fixing the islet pH_i at known values (ranging from 5.5 to 9, with 5–10 islets for each pH), using a KRBH medium containing 100 mM K⁺ and 20 mM nigericin to equilibrate the pH inside and outside cells.

All incubations were done in a 37°C water bath. Freshly isolated islets were pre-incubated for one hour in basal KRBH containing 2.8 mM glucose. Islets were then divided into groups and stimulated with different compounds as indicated in the results section, for one hour. Control group was maintained in basal glucose. At the end of the stimulation period, samples were collected for insulin measurement by radio-immuno-assay (performed by the DRTC Core facility at Vanderbilt University). Islet insulin content was measured after freezing islets overnight in 2% Triton-X. Insulin secretion was expressed as fractional release, i.e. the percentage of total insulin content released over the period of stimulation. In the experiments monitoring TDP, after pre-incubation islets were exposed to high (16.7 mM) glucose with or without intracellular acidification for 40 minutes, while the control group was exposed to basal glucose. Subsequently all groups were rested in basal glucose for 30 minutes, and stimulated with high glucose for one hour prior to collection of samples for insulin assay. The value n denotes the number of times each experiment was repeated using islets from different mice.

Experiments were performed using TPEM combined with techniques developed in our lab. a) Two-photon fluorescence imaging: Prior to imaging, cultured islets were maintained in RPMI medium containing 5 mM glucose for 24–48 hours. On the day of the experiment, RPMI medium was removed, and islets were washed and placed in basal KRBH medium. Two-photon imaging was done using a Plan-Apochromat 60 × 1.4 NA oil immersion lens of a LSM510 confocal laser-scanning microscope (Zeiss). Islets were kept at 37°C and high humidity using a temperature controlled stage and objective warmer (Zeiss). Sequential images of autofluorescence were collected using 710 nm excitation with a Coherent Mira laser tuned to 710 nm. Each image was collected with a single scan using slow speed (6.4 μs/pixel) and 0.1 μm/pixel. The laser power used provides ~3.5 mW to the surface, at which power no observable damage is caused to the islet 43 even after 60 minutes of continuous imaging 44. Non-descanned NAD(P)H fluorescence was collected through a custom 380- to 550-nm filter (Chroma). b) Standard Curve: Standard solutions of NADH bound to yeast alcohol dehydrogenase were imaged in deep well slides using live cell data collection settings. This solution closely resembles cellular NAD(P)H since a majority of NAD(P)H is enzyme bound in the cellular environment 45. NADH was titrated into a solution of ~100 mg/ml alcohol dehydrogenase, 10 mM ethanol, 100 mM isobutyramide, 6 mM semicarbazide, and 10 mM HEPES (pH 9.0). c) Image Analysis: Images were exported in tagged image file format (tiff) for analysis using MetaMorph software (version 5.0). Images were filtered using a 5 × 5 median filter. An intensity threshold was first set in the NAD(P)H image to isolate the bright mitochondria. In each image the maximum intensity from the mitochondrial regions was averaged. This average represents the best average intensity of the in focus mitochondria for the image. A threshold was set on the median filtered image to isolate cytoplasmic NAD(P)H intensity. The low end of this threshold excluded nuclear regions and the high end partially excluded brighter mitochondrial regions. A binary mask was made from the threshold. This mask was then eroded, dilated, and segmented to remove noise pixels in the background regions and generate a number of cytoplasmic regions. These regions were transferred to the median filtered image and the mean intensity of these regions was calculated as the average cytoplasmic NAD(P)H intensity.

Values are expressed as mean ± SEM. Groups were compared using paired Student's t test. In secretion studies, n denotes the number of times each experiment was repeated with islets from different mice. In imaging experiments for pH_i and NAD(P)H, n denotes the number of islets imaged for each condition.

SNARF5-AM loaded efficiently into a significant portion of the β cell-rich region of each islet, and emitted bright fluorescence when excited at 514 nm. Changes in the emission ratio 630/580 were directly proportional to changes in islet pH_i, indicating SNARF5-AM to be a suitable dye to monitor pH_i in mouse islets. (Fig. 1). The basal pH_i in islets (in KRBH with basal glucose) fell within the range of 6.9–7.2 with an average of 7.09 ± 0.01 (n = 72), and remained stable until additions/changes were made.

We tested several traditional methods of pH_i-alteration, including the addition of weak acids or amiloride derivatives (inhibitors of the Na⁺/H⁺ exchanger) for acidification, and addition of weak bases, removal of medium Cl^-, or raising the medium pH for intracellular alkalinization. The pH_i changes produced by weak acids or bases were sharp but very transient (data not shown), making these methods unsuitable for altering pH_i over a long period. In contrast, DMA produced a slow, steady and sustained decrease of pH_i, bringing it down by 0.2–1.0 pH units from baseline, with an average decrease of 0.5 ± 0.07 pH units (Fig. 2A). Removal of medium Cl^-, or raising the medium pH above 8.3, both produced sustained increases in pH_i, but the magnitude of the change was small, ranging from 0.1–0.4 units, with an average increase of 0.3 ± 0.05 and 0.24 ± 0.03, respectively (Fig. 2B &2C). Therefore, none of the traditional pH_i-increasing methods appeared suitable for producing a sufficient increase in pH_i for one hour. However, combination of the latter two methods, i.e. using a Cl^--free high pH medium, proved to be an effective technique for prolonged increase of pH_i by 0.3–0.7 units, with an average change of 0.5 ± 0.06. (Fig. 2D)

As has been shown before, TDP does not occur in mouse islets 151617. Remarkably, glucose induced a strong TDP response in mouse islets in the presence of intracellular acidification by DMA (Fig. 3). In other words, while the secretory response in mouse islets is not normally affected by a previous exposure to glucose, it was greatly magnified when the previous exposure to glucose was combined with lowered pH_i. It is noteworthy that the pH_i change caused by DMA does not persist once DMA is removed from the medium (data not shown), so that the enhancement of the subsequent secretory response to glucose is caused purely by a memory effect. Since lowering pH_i can induce/unmask a secretory function which is normally absent, it is very likely that pH_i has a strong influence on direct insulin secretion as well.

Previous studies have shown that GSIS is enhanced by intracellular acidification and inhibited by intracellular alkalinization 23242526272829303132. Our preliminary studies confirmed this, and showed that even weak methods of alkalinization (such as removal of medium Cl^- or raising the medium pH above 8.3) significantly inhibit GSIS (Fig. 4A).

Among the mitochondrial secretagogues tested, leucine, αKIC (metabolic product of leucine) and BCH (non-metabolizable analog of leucine) all produced significant insulin secretion, to a lesser degree than glucose. As expected, insulin secretion by all these agents showed a strong correlation with pH_i. Decrease of pH_i by DMA caused a marked increase in the insulin secretion, making the magnitude of the secretory response close to that of glucose. Insulin release by all three agents was significantly inhibited by a strong increase in pH_i, as produced by a Cl^--free high pH medium (Fig. 4B,4C &4D).

Interestingly, weaker methods of alkalinization were not sufficient to inhibit the insulin release by leucine, KIC and BCH, unlike with glucose (data not shown). This difference can be explained by the direct effect of each nutrient on intracellular pH. As shown in Table 1, glucose produced a significant increase in pH_i. This was further increased even by weaker methods of alkalinization, so that high glucose in a high pH medium (Table 1) or Cl^--free medium (data not shown) resulted in a prolonged increase of pH_i to 7.5 or higher. A high pH medium (Table 1) or Cl^--free medium (data not shown) alone was not adequate to produce such a rise in pH_i, with or without the presence of other fuels such as leucine, BCH or αKIC. Therefore, while any method of alkalinization combined with glucose can drive the pH_i above the range favorable for secretion, weaker methods of alkalinization still maintain the pH_i within this favorable range in the presence of other mitochondrial fuels (Table 1). Insulin release by these compounds could be inhibited only when pH_i was driven above 7.5 by a Cl^--free high pH medium (Fig. 4B,4C &4D).

The inability of pyruvate to stimulate insulin release is puzzling, and has been widely studied 394647. We used the methylated form of pyruvate (MP), which is known to penetrate cellular membranes better than pyruvate 48495051, as one of the mitochondrial stimulants. In our hands, MP did not stimulate insulin secretion, with or without acidification by DMA (Fig. 5A). Interestingly, MP caused a dramatic and prolonged decrease in pH_i (ranging from 0.7–1.7 units from baseline, with an average decrease of 1.0 ± 0.08 pH units), providing a possible explanation for its lack of secretion (Fig. 5B). If the lack of secretion by MP is due to inappropriately low pH_i, then a forced increase of pH_i should allow it to stimulate insulin release. While the weaker methods of alkalinization were not adequate to reverse the acidifying effect of MP, a Cl^--free high pH medium did reverse it, bringing the pH back to the range of 6.4–7.0 (Fig. 5C,5D &5E). As expected, reversing the acidifying effect of MP by a Cl^--free high pH medium did allow MP to stimulate significant insulin secretion. (Fig. 5F).

High K⁺ causes insulin exocytosis via depolarization-induced Ca²⁺-influx, and does not involve nutrient metabolism or ATP production. K⁺-induced insulin secretion was slightly but not significantly enhanced by intracellular acidification, and was not affected by alkalinization (Fig. 6A).

Glucose is converted to a non-miotchondrial secretagogue by combining with the mitochondrial inhibitors AOA and CHC 39. As expected, the presence of these inhibitors caused a significant reduction in the insulin response by glucose. Remarkably, intracellular acidification by DMA not only reversed this inhibition, but enhanced the insulin secretion further than that produced by glucose alone (Fig. 6B). This suggests that the effect of pH_i is not limited to mitochondrially-derived signals, and that extra-mitochondrial signals alone can induce significant insulin secretion when pH_i is within the favorable range.

To explore the mechanisms whereby DMA reversed the inhibition of GSIS by AOA and CHC, we monitored insulin secretion and cellular NAD(P)H response in the presence/absence of DMA and the inhibitors. While DMA enables glucose to stimulate insulin release in the presence of inhibitors (Fig. 6B), DMA alone does not stimulate insulin secretion (Fig. 4A). As shown in fig. 7, however, DMA did increase whole cell NAD(P)H levels in all situations tested, including basal or high glucose with and without inhibitors. It is interesting that DMA increased NAD(P)H at basal glucose levels (Open bars: no additions vs. DMA, and AOA/CHC vs. AOA/CHC/DMA). This is likely due to an increase in NADPH rather than NADH, based on the fact that DMA alone does not cause insulin secretion. High glucose induced similar increases in NAD(P)H in the absence or presence of DMA, suggesting that DMA does not stimulate glucose utilization beyond normal. However, when glucose metabolism is inhibited with AOA and CHC, DMA treatment allowed glucose to induce similar changes as those observed in the absence of the inhibitors. This suggests that DMA treatment somehow enables the full utilization of glucose. To further investigate the metabolic effect of low pH_i, we looked at the effect of DMA on cytosolic and mitochondrial NAD(P)H levels in the presence and absence of nutrients whose secretory ability is enhanced at low pH_i. Both in the presence and absence of nutrients (16.7 mM glucose or 20 mM αKIC), DMA produced a dramatic increase in cytoplasmic NAD(P)H levels (Fig. 8A), and a smaller but significant increase in the mitochondrial NAD(P)H levels (Fig. 8B). Again, glucose-induced increases in the NAD(P)H response (both cytoplasmic and mitochondrial) were of similar magnitude in the absence or presence of DMA.