Glucose-Stimulated Insulin Secretion Fundamentally Requires H₂O₂ Signaling by NADPH Oxidase 4

Changes in ROS/H₂O₂ during GSIS in INS-1E cells. INS-1E cells were transfected with scrambled siRNA (ntg cells [black]) or with two NOX4 siRNAs (red) under standard cell cultivation (11 mmol/L glucose [Glc]) or upon preincubation with 3 mmol/L Glc for 15 h. Quantifications in relative units (normalized) before and after Glc was increased to 25 mmol/L (gray bars [“+Glc”]) are shown, as are mean fluorescence (F¹¹₀) and fluorescence rate (J_c¹¹₀) for standard cultivation with 11 mmol/L Glc before GSIS. A: Cytosolic ROS release was monitored by using the fluorescence of dihydro-dichlorofluorescein diacetate (DCF) (excitation wavelength, 480 nm; emission wavelength, 500–550 nm) in a Shimadzu RF-5301PC spectrofluorophotometer (see exemplar traces in this panel). B: Quantification of rates of increased fluorescence of DCF normalized to that of the controls (% wt). C: Changes in static DCF fluorescence values in the presence of 10 μmol/L DPI (D), 0.5 μmol/L stigmatellin (S), or 10 μmol/L DPI + 0.5 μmol/L stigmatellin (DS), normalized to controls at 11 mmol/L glucose. D, E, G, and H: Changes (normalized to controls at 11 mmol/L glucose) in the static fluorescence of HSP33-FRET (D, E, and H) and J_c rates of increased HSP33-FRET fluorescence, representing cytosolic H₂O₂ release (cH₂O₂) (G), monitored at an excitation wavelength of 430 nm and an emission wavelength of 470–535 nm. Where indicated, 1 mmol/L 6-aminonicotinamide (6AN) or 40 μmol/L oxythiamine was present. F: cH₂O₂ release rates (J_c) monitored with a HyPer-C (excitation wavelength, 488 nm; emission wavelength, 488–516 nm); rates were normalized to that of the controls at 11 mmol/L Glc. **P < 0.05; ***P < 0.001, ANOVA (Tukey test) (N = 3 and n = 10 if not stated otherwise); n = 13 in C; n = 14 in D (n = 10 after GSIS). (For the Holm-Sidak test, see the Supplementary Appendix). I: Superoxide release into the mitochondrial matrix, monitored by MitoSOX Red under confocal microscopy. Fluorescence rates (J_m) are proportional to the surplus superoxide released into the matrix (36). J¹¹_m0, mean of the standard ntg cell cultivation before GSIS at 11 mmol/L Glc (N = 3 and n = 11; n = 10 in 3 mmol/L Glc for 15 h). **P < 0.05; ***P < 0.001, ANOVA (Tukey test). (For the Holm-Sidak test, see the Supplementary Appendix).

GSIS suppression, IGT, and peripheral insulin resistance (IR) in NOX4KO mice (orange) and NOX4βKO mice (red). Insulin released to blood circulation (A, B, E, F, and I) and glycemia (C, D, G, H, and J) were both evaluated in blood from the ocular plexus after i.p. injection of a glucose bolus (1 mg/g body weight) in mice that had been deprived of food overnight (wild-type backcrossed mice [9 males, 10 females] [black] or NOX4KO mice [20 males, 14 females]) (A and C) and NOX4^Flox/Flox (16 males, 14 females) (black) or NOX4βKO mice (19 males, 18 females) (B and D). E–H: Glibenclamide (5 ng/g body weight) without glucose was also injected i.p. (12 males and 12 females for each strain). I and J: Likewise, the corresponding tests were performed in NOX2KO mice (two males, one female; green) and mice backcrossed from them (three males, two females; black). Areas under the curve (AUCs) constructed from means (or the mean ± SD for controls) of insulin release are 38–53% for NOX4KO vs. wild-type backcrossed mice (138–162% for glycemia) and 12–25% for NOX4βKO mice vs. NOX4^Flox/Flox mice (141–184% for glycemia). After glibenclamide was administered, AUCs were 126–97% (up to the mean ± SD for knockout mice) for NOX4KO vs. wild-type backcrossed mice (79–135% for glycemia) and 43–81% (up to the mean ± SD for knockout mice) for NOX4βKO mice vs. NOX4^Flox/Flox mice (634–670% for glycemia). GSIS in NOX2KO mice (green) exhibited an AUC that was 89–93% of that for the corresponding backcrossed mice (108–105% for glycemia). Samples from two or three time points from each mouse were used to estimate insulin and glycemia. Time dependencies were constructed from numerous groups of mice (comprising different littermates, both males and females; 19–37 mice were in each group; 3 for NOX2KO, 5 for its backcrossed mice), covering all the required time points. The different littermates are indicated with different symbols. This setup represents the only way to perform fast sampling of insulin release by using the most sensitive insulin ELISA kit (Mercodia).*P < 0.1; **P < 0.02; ***P < 0.001 or <0.002 (Student t test; n = 5–20 data pairs at each time point). (For notable P values, see the Supplementary Appendix.) K and L: Peripheral IR indicated by inhibited ¹⁴C-glucose uptake into the glycogen of the diaphragm (K) and lipids of epidydimal adipose tissue (L). ***P < 0.001, ANOVA (Tukey test) (N = 7; always male mice). (For original data, see Supplementary Fig. 10.) Note that the Holm-Sidak test yielded P values of 2.3 × 10⁻⁷ and 2.5 × 10⁻¹⁰ in K and values of 3.3 × 10⁻⁵ and 0.0036 in L.

GSIS suppression in PIs isolated from NOX4KO and NOX4βKO mice, NOX4-dependent H₂O₂ cytosolic release, and rescue of GSIS with Nox4 overexpression or H₂O₂. Colors of bars and symbols are as presented in Fig. 3. A–H: Perifusion of PIs. Mean time courses (N = 3 isolations each from two males and two females, or the typical course from a single isolation) of insulin release in PIs (preincubated in 2.5 mmol/L glucose) with initial additions of 25 mmol/L (A–D) or 5 mmol/L (G and H) glucose; in some cases, 25 μmol/L palmitic acid and 25 μmol/L oleic acid were coinjected within a medium containing 2 mmol/L glutamine (G and H). When indicated, the NOX4-selective inhibitor GKT-137831 (5 μmol/L; magenta) was added to control PIs, or 100 μmol/L glibenclamide was added. The marks at 60 min indicate the levels when 30 mmol/L KCl was added at the end of the perfusate. Until 18 min, areas under the curve (AUCs) were 38% (85% with fatty acids) in PIs from NOX4KO mice and 12% (70% with fatty acids) in PIs from NOX4βKO mice. AUCs obtained from the data were related between PIs from NOX4KO mice and those from NOX4βKO mice, as follows: 38% until 18 min (A), 12% until 18 min (B), 31% until 14 min (74% for glibenclamide phase) (C), 18% until 14 min (73% for glibenclamide phase) (D), 113% (E), 55% (F), 85% (G), and 96% (H); *P < 0.1; **P < 0.05; ***P < 0.001, ANOVA (Tukey test) (N = 3; n = 5–7). With the Holm-Sidak test, notable P values varied. In A, P < 0.001 at 4 min, P = 0.006 at 6 min, P = 0.011 at 8 min, and P = 0.047 at 10 min. In B, P < 0.001 at 4, 6, and 8 min; P = 0.003 at 10 min; and P = 0.009 at 12 or 14 min. In D, P = 0.001 at 4 or 6 min and P = 0.004 at 8 min (otherwise see the Supplementary Appendix). I–K: Cytosolic ROS release monitored in PIs with DCF, as described in Fig. 2. Both exemplar traces (I and J) and quantification of rates (K) are shown. ***P < 0.001, ANOVA (Tukey test) (n = 10). L: Rescue experiment in PIs. Lentiviral NOX4 overexpression (N = 3; blue) in PIs of NOX4βKO and NOX4^Flox/Flox mice (NOX4^+/+ [dashed bar]) vs. in “empty” green fluorescent protein reporter vector expression (GFP). Obtained statistics: magenta (yellow) points. ***P = 0.007, ANOVA (Tukey test). For the Holm-Sidak test, notable values were P = 0.0016, GFP vs. NOX4βKO NOX, and P = 0.00032, NOX4βKO vs. NOX4βKO rescue. The identity of values was indicated by P = 0.321 for NOX4^+/+ and NOX4βKO rescue, and P = 0.222 for GFP and NOX4βKO rescue. (Otherwise, see the Supplementary Appendix.) M and N: Rescue with H₂O₂ in PIs from NOX4KO mice. Rescue was evaluated as in A and B, but the perfusate contained 100 μmol/L H₂O₂ for the first 10 min after glucose was added. Typical runs are shown (N = 3). Gray traces represent only glucose, without H₂O₂. M and N: AUCs of H₂O₂ rescue in PIs from NOX4KO (NOX4βKO) mice vs. H₂O₂ plus glucose or only glucose in wild-type backcrossed (NOX4^Flox/Flox) mice were 107% or 157% until 10 min (M) and 66% or 90% until 10 min (N). O: Glucose-induced increase in total ATP in PIs. Values were normalized to the initial levels at 3 mmol/L glucose.

Dependence of K_ATP channel status on NOX4. The K_ATP channel opened upon GSIS, as indicated by the patch clamp in cell-attached mode. Negative potentials refer to the interior of the plasma membrane, and inward currents deflect downward. A: In ntg cells, the K_ATP channel was fully open at L3 and partially open in states L1 and L2. B: In ntg cells with glucose, there was an increased number of L1 and L2 states but a reduced number of L3 occurrences. C: NOX4-silenced INS-1E cells (NOX4KO) show insignificantly higher probabilities that the K_ATP channel will be open (NP_o) than do ntg cells. D: K_ATP activity is slightly reduced in silenced cells treated with glucose. E: The i-E relation for the L3 state of the K_ATP channel is unchanged, demonstrating constant single-channel conductance γ. F: The probability that the channel would be open (NP_o) was calculated for the time the K_ATP channel spent in open states (t_o) divided by the total recording time T (usually 1 min) by using the single-channel search mode in Clampfit software version 9.2. When there were indications that the patch contained more than one channel, the program used NP_o = T_o/(T_o + T_c), where N is the minimum number of channels, and T_o = ΣLt_o, where L is the number of openings. For comparison, NP_o values were normalized to the NP_o value for ntg cells, and the different P_o values were normalized to T_o under those same conditions. The resting potential of the ntg cells was −70 mV before and −40 mV after glucose was added. This demonstrated the depolarization of the plasma membrane by glucose in ntg cells. All the open states were, again, completely closed by glibenclamide. Once the glucose was applied, the excision of the patch to inside-out mode allowed K_ATP activity to recover. The recovered K_ATP activity was inhibited after the addition of glibenclamide. G: L1–L3 contribution to total P_o. Glucose reduced the percentage contribution of L3 and L2 states to the benefit of L1, and it did so strongly in ntg cells. In F, ANOVA (N = 5; n = 5–7) yielded notable P values according to the Holm-Sidak test: P = 0.09 (0.31 for Tukey test) for ntg vs. siRNA NOX4, indicating equal initial NP_o values before glucose was added; and P = 0.00028 (**P = 0.002, Tukey test), indicating significantly high NP_o values after glucose was added in siRNA NOX4 samples. For ntg cells before and after glucose, P = 6.4 × 10⁻⁶ indicates significance.

GSIS or RSIS vs. the status of the K_ATP channel in INS-1E cells. A–C: Normalized rates of GSIS or RSIS (H₂O₂) for the indicated agents (C). The RSIS time courses after 98 μmol/L H₂O₂ (with or without 25 mmol/L glucose [Glc] added at the beginning) (A); related quantifications of insulin released after 60 min (B). In B, **P < 0.05, ANOVA (Tukey test; n = 7). In C, ***P < 0.001 (N > 3; n = 5–7) relative to GSIS in respective ntg controls. (For the Holm-Sidak test see the Supplementary Appendix.) D, E, and G: K_ATP status as reflected by Tl⁺ influx into ntg control (black bars), NOX4-silenced (NOX4 siRNA; red bars), and catalase-overexpressing INS-1E cells, with 25 mmol/L Glc added or after direct addition of H₂O₂ (98 and 196 μmol/L, as indicated), without or with 50 μmol/L diazoxide (white bars), 1.3 μg/mL oligomycin (H₂O₂Olig.), or only oligomycin (Olig.); or with 15 mmol/L OIC or 20 μmol/L palmitic acid (G): ***P < 0.001, ANOVA (Tukey test) (N = 3–5; n = 10; n = 5 with reagents; related to GSIS in D). (For the Holm-Sidak test see the Supplementary Appendix.) F: Plasma membrane depolarization, as monitored with PPMP 5 min after 25 mmol/L Glc was added; values were normalized to those in ntg control cells. H₂O₂ after glc, 25 mmol/L Glc is already present; H₂O₂ before glc, addition of H₂O₂ alone (98 μmol/L). Fluorescence monitoring of plasma membrane potential, E_m, was conducted with the plasma membrane potential indicator (PMPI) (component-A, FLIPR Membrane-Potential Assay Kit; Molecular Devices, Sunnyvale, CA), added just 20 min before imaging. Cells were excited at 514 nm, with an emission wavelength of 525 nm. ***P < 0.001 vs. ntg, ANOVA [Tukey test]; n = 5, N = 2). (For the Holm-Sidak test, see the Supplementary Appendix). H–M: GSIS or RSIS vs. the status of the Ca_L in INS-1E cells. H–J: Ca²⁺ influx, microscopically monitored with Fura-2. Black and light blue traces represent the ntg cells (controls); red and dark blue traces represent NOX4KO cells. The arrows indicate the points where glucose (25 mmol/L) or H₂O₂ (98 μmol/L) were added, without or with 50 μmol/L cromakalim. K–M: Occurrence of the peak values each minute (K), no peaks (L), or the first peak (M), as evaluated from Fura-2 assays (50–100 peaks; N = 2).

NOX4-independent, oxoacid-stimulated insulin secretion requires a mitochondrial redox burst plus ATP. Throughout these graphs, NOX4-silenced INS-1E cells or NOX4βKO mice are indicated with the color red; NOX4KO mice, orange; backcrossed mice as the control for NOX4KO mice, black; NOX4^Flox/Flox mice as the control for NOX4βKO mice, gray; BCKDH-silenced INS-1E cells, cyan; 500 nmol/L SkQ1, yellow. A and B: Insulin secretion responding to 15 mmol/L OIC in INS-1E cells, with or without 1 nmol/L SkQ1 or 4 mmol/L aminooxyacetate (AOA). C: Quantification of the rate of insulin secretion. ***P < 0.001 vs. ntg GSIS values, ANOVA (Tukey test) (N > 4, n = 5–7). (For the Holm-Sidak test, see the Supplementary Appendix.) E–J: Insulin in the blood (E and F) and glycemia (G and H) after i.p. injection of 1 mg OIC/g body weight (∼154 μmol per mouse), or in PIs perifused with 15 mmol/L OIC (2.5 mmol/L glucose); for simplicity, a typical example is shown from N = 3 time courses/isolations for each situation (I and J). Areas under the curves (AUC) constructed from means of insulin release were 106% until 15 min for NOX4KO vs. wild-type backcrossed mice in E (177% for glycemia in G), and 58–77% (between means, up to between the mean ± SD) for NOX4βKO mice vs. NOX4^Flox/Flox mice in F (232% for glycemia in H). For PIs, AUCs were related to proper control: 112% until 10 min and 60% until 60 min (I) or 47% until 10 min and 54% until 60 min (J). Here, 15 male and 20 female NOX4KO mice and 15 male and 20 female backcrossed mice were used, as were or 57 male and 32 female NOX4βKO mice and 25 male and 36 female NOX4^Flox/Flox mice (though data from only 35 are displayed). **P < 0.05, ***P < 0.001, Student t test (n = 8–15 vs. n = 8–15 data pairs for NOX4KO mice and n = 13–21 vs. n = 13–21 data pairs for NOX4βKO mice at each time point). The different littermate groups are indicated by the different symbols. For notable P values, see the Supplementary Appendix. D: Superoxide release into the matrix with OIC (see L), normalized to rates of superoxide release in cells preincubated with 3 mmol/L glucose with nothing else added (J_m³₀). Note that the addition of 25 mmol/L glucose causes the J_m rate to decrease (Fig. 2I), whereas the ETF:QOR contribution of OIC metabolism, which is dependent on BCKDH, causes the superoxide release to be sensitive to the mitochondrial matrix antioxidant SkQ1. ***P < 0.001 vs. ntg GSIS values yielded, ANOVA (Tukey test) (N > 3; n = 5–7) . (For the Holm-Sidak test, see the Supplementary Appendix.) K: Typical insulin release with 1 nmol/L SkQ1 (yellow) after the addition of 25 mmol/L glucose. L: The chemical structure of OIC. M: OIC-stimulated insulin secretion, which is predominantly independent from NOX4 and uses ROS generated by mitochondrial ETF:QOR. BCAT, branched-chain α-ketoacid amino transferase; IVD, isovaleryl-CoA dehydrogenase; MCC, methylcrotonyl-CoA carboxylase; MGCoAH, methyl-glutoconyl-CoA hydratase; HMGCoAL, 3-hydroxy-3-methylglutaryl-CoA lyase.