Insulin secretion stimulated by l-arginine and its metabolite l-ornithine depends on Gα_i2

Abstract

Bordetella pertussis toxin (PTx), also known as islet-activating protein, induces insulin secretion by ADP-ribosylation of inhibitory G proteins. PTx-induced insulin secretion may result either from inactivation of Gα_o proteins or from combined inactivation of Gα_o, Gα_i1, Gα_i2, and Gα_i3 isoforms. However, the specific role of Gα_i2 in pancreatic β-cells still remains unknown. In global (Gα_i2^−/−) and β-cell-specific (Gα_i2^βcko) gene-targeted Gα_i2 mouse models, we studied glucose homeostasis and islet functions. Insulin secretion experiments and intracellular Ca²⁺ measurements were used to characterize Gα_i2 function in vitro. Gα_i2^−/− and Gα_i2^βcko mice showed an unexpected metabolic phenotype, i.e., significantly lower plasma insulin levels upon intraperitoneal glucose challenge in Gα_i2^−/− and Gα_i2^βcko mice, whereas plasma glucose concentrations were unchanged in Gα_i2^−/− but significantly increased in Gα_i2^βcko mice. These findings indicate a novel albeit unexpected role for Gα_i2 in the expression, turnover, and/or release of insulin from islets. Detection of insulin secretion in isolated islets did not show differences in response to high (16 mM) glucose concentrations between control and β-cell-specific Gα_i2-deficient mice. In contrast, the two- to threefold increase in insulin secretion evoked by l-arginine or l-ornithine (in the presence of 16 mM glucose) was significantly reduced in islets lacking Gα_i2. In accord with a reduced level of insulin secretion, intracellular calcium concentrations induced by the agonistic amino acid l-arginine did not reach control levels in β-cells. The presented analysis of gene-targeted mice provides novel insights in the role of β-cell Gα_i2 showing that amino acid-induced insulin-release depends on Gα_i2.

glucose is the principal stimulator of insulin secretion from pancreatic β-cells. Together with regulators such as other nutrients or hormones, it adjusts insulin secretion according to physiological demands. A disruption of this tightly controlled process can lead to diabetes mellitus and its comorbidities. Physiological regulators of insulin release include not only glucose but also other nutrients such as free fatty acids or amino acids on the one hand, and hormones including glucagon and norepinephrine on the other hand. They all have in common that they signal via seven-transmembrane G protein-coupled receptors (GPCR) present on the surface of pancreatic β-cells (1, 28, 33). Upon ligand binding, GPCRs activate heterotrimeric G proteins, thereby modulating the activity of cellular effectors. It is widely accepted that insulin release from pancreatic β-cells can be triggered via Gα_s- and/or G_q/G₁₁-dependent mechanisms (2, 41). Studies in isolated systems and in animals, using pertussis toxin (PTx), also known as islet-activating protein, suggest Gα_i/Gα_o-dependent signaling as being important for inhibition of insulin secretion (11, 14, 15). PTx specifically catalyzes the ADP-ribosylation of a cysteine residue located four residues from the carboxyl terminus of the α-subunits of Gα_i/Gα_o proteins. Thus, PTx treatment disrupts GPCR stimulation of the G protein and thereby disconnects it from signal transduction. The α-subunits of the PTx-sensitive G proteins include three Gα_i (Gα_i1, Gα_i2, and Gα_i3) proteins, which are the product of three different genes, and two Gα_o splice variants i.e., Gα_o1 and Gα_o2 (3). Gα_i/Gα_o proteins show a sequence identity of up to 95% with overall sequence homology being more pronounced among the three Gα_i isoforms (37). Due to this high homology, it has been suggested that all Gα_i and Gα_o proteins serve the same functions; namely, they are activated by the same or a similar set of GPCRs and appear to signal to a fully overlapping set of effectors. However, combined analysis of global and islet cell-specific Gα_o1- and/or Gα_o2-deficient mouse lines and PTx treatment identified Gα_o2 as the main inhibitory G protein related to insulin release (32, 34, 42). Glucose-induced insulin secretion (34) and insulin containing vesicle docking (42) were increased in the absence of Gα_o2. Moreover, PTx treatment of Gα_o-deficient mice had a negligible effect on the already disarranged insulin secretion (42). These findings indicate that Gα_o proteins are the major isoform responsible for PTx-sensitive inhibition of insulin secretion. Additional evidence suggests that Gα_o2 mediates the somatostatin (SST)- and galanin-induced receptor-to-effector signal transduction in β-cells (32). Together, these findings support a prominent role for Gα_o2 as the major regulator in GPCR-induced inhibition of insulin secretion. It is presently largely unclear whether the PTx-sensitive Gα_i2 isoform expressed in islets or elsewhere, such as intestine, skeletal muscle, or adipose tissue is of any significance for insulin secretion in vitro and/or glucose homeostasis in vivo. To elucidate the biological role and gain insight into the cellular mechanisms of Gα_i2 protein function in pancreatic β-cells, we examined global and β-cell-specific Gα_i2-deficient mice. We show that Gα_i2 mediates l-arginine-induced insulin secretion by modulation of corresponding changes in [Ca²⁺]_i that trigger exocytosis. Decreased plasma insulin levels in Gα_i2-targeted animals leading to an impaired glucose tolerance supported these in vitro findings. These data strongly promote and unexpectedly stress a stimulatory role for Gα_i2 in insulin secretion, showing that although PTx inhibits Gα_o and Gα_i2 proteins these two isoforms have distinct and opposite roles in the regulation of insulin secretion. Whereas Gα_o inhibits, Gα_i2 stimulates insulin secretion from β-cells.

MATERIALS AND METHODS

Experimental animals.

The generation of the global Gα_i2-deficient mice (Gα_i2^−/−) has been described previously (29). To prolong life expectancy of the Gα_i2^−/− animals, mice were bred and kept in individually ventilated cages (IVC) under specific-pathogen-free conditions (39) and had free access to water and standard chow. β-Cell-specific deletion of Gα_i2 (Gα_i2^βcko) was achieved by crossing the floxed Gα_i2 mouse line (27) and the Rip-Cre^+/tg mouse line (12), both on a C57BL/6N background. The Rip-Cre^+/tg mouse line was a kind gift from Pedro Luis Herrera, Geneva. Gα_i2^βcko and their littermate controls (ctrl) were kept under specific-pathogen-free conditions with free access to water and standard chow. For all experiments, gene-targeted Gα_i2 animals were compared with their littermate wild types (WT) or controls (genotype: Gα_i2^+/+;Rip-Cre^+/tg or Gα_i2^+/fl;Rip-Cre^+/tg) on a C57BL/6N background (age 6–24 wk). All animals analyzed within this study were maintained and bred in the animal facility of the Institute of Experimental and Clinical Pharmacology and Toxicology, Eberhard Karls University, Tübingen. All experimental procedures were approved by the local government's Committee on Animal Care and Welfare Tübingen.

Isolation of pancreatic islets.

Islets were isolated according to procedure in Ref. 20. Upon a retrograde injection of 2 ml of collagenase solution (1.9 U/ml collagenase; Sigma Aldrich, Munich, Germany) in 1× HBSS (Life Technologies, Darmstadt, Germany) via the bile duct, fully perfused pancreata were incubated for 10 min at 37°C. Islets were picked using a 200-μl pipette during several washing steps.

Whole pancreatic insulin content.

Pancreata were removed and weighed. After a homogenization step in 7 ml of ice-cold 1.5% HCl in 70% ethanol, pancreata were incubated in acid ethanol for 24 h at 4°C. Insulin content was determined in the supernatant with the commercially available ultrasensitive insulin ELISA method (Mercodia, Uppsala, Sweden). Pancreatic insulin content was normalized to the weight of the pancreas.

Islet morphology.

Whole pancreatic tissue was fixed in 4% formaldehyde for 24 h and embedded in paraffin at a stretch according to standard procedures, and longitudinal 6-μm serial sections were performed. Every 250 μm, pancreata were stained with hematoxylin and eosin, and every single islet within the stained pancreatic slice was quantified. Average islet and cell size was determined by hand, whereas the number of nuclei were calculated with the AxioVision software (Zeiss, Jena, Germany). Mean values derived from each individual mouse were used to calculate means ± SE of three mice in each group.

In vitro experiments with isolated islets.

The effects of treatments on insulin secretion were assessed using static incubation experiments (20). Overnight-cultured islets [RPMI 1640 (GE Healthcare Europe, Freiburg, Germany), 10% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin] were grouped as batches of 10 islets and incubated in Krebs-buffered HEPES saline (120 mM NaCl, 4.7 mM KCl, 2.5 mM CaCl₂, 1 mM KH₂PO₄, 1.2 mM MgSO₄, 10 mM HEPES, 20 mM NaHCO₃, 0.5 mg/ml BSA, pH 7.4) containing 3 mM glucose in humidified 5% CO₂ at 37°C for 30 min. Subsequently, the supernatant was discarded, and fresh Krebs-buffered HEPES saline was added supplemented with either 3 or 16 mM glucose with or without l-arginine (10 mM), l-ornithine (10 mM), l-leucine (100 μM), palmitate (0.2 mM), clonidine (1 μM), SST (0.1 μM), or KCl (30 mM). The islets were incubated in humidified 5% CO₂ air at 37°C for 1 h. Afterward, islets were treated with ice-cold acid ethanol (1.5% HCl, 70% ethanol) for 1 h. For each condition, aforementioned above 3 and 16 mM glucose was used to test the quality of islet preparation. The amount of secreted insulin in the supernatant as well as the whole insulin content was determined with a commercially available ultrasensitive insulin ELISA method (DRG Diagnostics, Marburg, Germany). Islet DNA was isolated with the Isolate II Genomic DNA kit (Bioline, Luckenwalde, Germany), and the islet DNA content was measured (30). Based on the identical DNA and whole insulin contents of Gα_i2^βcko and ctrl islets, the amounts of secreted insulin were indicated % of insulin content.

Immunoblot analysis.

Freshly isolated islets were homogenized in 0.8 μl of protein lysis buffer per islet. The protein extractions were performed on pooled islets isolated from a single animal. To achieve electrophoretical separation of Gα_i isoforms, separation was performed in gels containing 6 M urea. The proteins were visualized by immunodetection using the following primary antibodies described elsewhere (7, 10, 21, 39): rabbit anti-Gα_i2 (1:8,000 islets, 1:20,000 hypothalamus), rabbit anti-Gα_i3 (1:5,000), rabbit anti-Gα_o1/o2 (1:2,000), and rabbit anti-Gβ_common (1:9,000). Antibodies against Gα_{s(short)/s(long)} (1:1,000; SC-383) and Gα_q (1:1,000; SC-393) were purchased from Cell signaling. Equal loading was verified with antibodies against mouse anti-α-tubulin (1:4,000, Sigma-Aldrich) or mouse anti-β-actin (1:4,000 islets, 1:40,000 hypothalamus; Sigma-Aldrich). The protein levels of Gα_i2, Gα_i3, Gα_o, Gα_q, Gα_s, Gβ₁, and Gβ₂ were quantified using densitometric analysis software (Image Lab; Bio-Rad, Gräfelfing, Germany) and were normalized to the α-tubulin or β-actin levels of the same samples.

Immunofluorescence.

Immunodetection was performed on serial cryosections (8 μm) of pancreata fixed in 4% paraformaldehyde. For antigen retrieval, slices were treated with 0.3% Triton X-100-PBS for 30 min. To avoid unspecific antibody binding, slices were incubated in 5% normal goat serum (NGS)-PBS prior to incubations in primary antibody dilutions. Primary antibodies used were specific for rabbit anti-Gα_i2 (1:1,500), mouse anti-glucagon (1:5,000; Sigma-Aldrich, Hamburg, Germany), and guinea pig anti-insulin (1:5,000; DAKO, Hamburg, Germany) and diluted in 1.5% NGS-PBS. Binding of primary antibodies was performed at 4°C overnight. The complexes were detected with secondary antibodies conjugated to fluorescent dyes (1:200 dilution in PBS, Invitrogen). Slices were embedded in Roti-Mount-Fluoro-Dapi (Carl Roth, Karlsruhe, Germany). Fluorescence was visualized using either the Zeiss laser scanning microscope LSM 510 or the Zeiss Axio Image.M2 microscope (Zeiss, Jena, Germany).

Glucose tolerance tests.

Mice were fasted overnight, followed by an intraperitoneal (ip) injection of glucose (2 mg/g body wt). Blood samples were collected via the tail vein, and blood glucose levels were measured using a Contour glucometer (Bayer, Leverkusen, Germany).

l-Ornithine test.

Mice were fasted overnight but supplemented with 1 mM l-ornithine in their drinking water to ensure sufficient high basal l-ornithine levels in the plasma. Mice received 100 μl of a 1.32 mg/ml l-ornithine solution by ip injection, and blood samples were collected via tail vein before administration and after 1 min to determine plasma insulin levels with the commercially available ultrasensitive insulin ELISA method (see above).

Insulin tolerance test.

Mice were fasted for 4 h, followed by an ip injection of insulin (1 mU/g body wt). Blood glucose levels were measured using a Contour glucometer (Bayer).

Plasma analysis.

Plasma levels of insulin were measured using a commercially available insulin ELISA method (see above), whereas plasma C-peptide levels were determined with the 96-well fluorescent Milliplex immunoassay (Millipore, Darmstadt, Germany).

Ca²⁺ imaging.

Intracellular calcium measurements were performed as previously described (16). Isolated islets were dispersed by trypsin-EDTA digestion, plated on glass coverslides, and allowed to recover for 4 h in humidified 5% CO₂ air at in 37°C. Cells were loaded with Fura 2-AM (5 μM) 30 min prior to calcium measurements. Changes of intracellular calcium concentrations [Ca²⁺]_i were recorded with a Till Photonics Oligochrome V (FEI, Gräfeling, Germany). After 4 min of baseline recordings in the presence of 5 mM glucose, dispersed islet cells were treated with 16 mM glucose, 10 mM l-arginine, 10 mM d-arginine, or a combination of 16 mM glucose and l-arginine for 10 min. Maximum depolarization of β-cells was evaluated by the use of 30 mM KCl. Emitted fluorescence was excited at 340 and 380 nm and measured at 520 nm, and the ratio 340/380 was calculated. Cells were identified as β-cells, when being inactive at 5 mM glucose, with stable baseline recordings, and an obvious response to the applied stimuli (e.g., 16 mM glucose, 10 mM l-arginine). Cells without Ca²⁺ responses to the respective stimuli were excluded from the measurements and analysis.

Statistical analysis.

Data are presented as means ± SE of individual data points. Data were analyzed using Student's t-test for unpaired groups. P < 0.05 was considered statistically significant.

RESULTS

Characterization of β-cell-specific Gα_i2-deficient mice.

The generation and characterization of global Gα_i2-deficient (Gα_i2^−/−) mice has been described (27, 29, 39). The β-cell-specific Gα_i2-deficient mice, termed Gα_i2^βcko (genotype: Gα_i2^fl/fl;Rip-Cre^+/tg), were generated by crossing Rip-Cre mice (12) with mice carrying floxed Gα_i2 alleles (27). The resulting progeny, i.e., the different genotypes, were born at expected Mendelian ratios and were viable and fertile. Since the expression of the Rip-Cre transgene has been reported to have metabolic effects per se (19), we used Rip-Cre-positive littermates throughout this study as controls (ctrl genotype: Gα_i2^+/+;Rip-Cre^+/tg or Gα_i2^+/fl;Rip-Cre^+/tg) for experiments including Gα_i2^βcko mice after having confirmed that heterozygous mice, e.g., Gα_i2^+/fl;Rip-Cre^+/tg mice were indistinguishable from Gα_i2^+/+;Rip-Cre^+/tg animals. Recombination was validated by PCR analysis of genomic DNA (Fig. 1, A and B). Whereas DNA isolated from hearts and islets from ctrl animals as well as cardiac DNA of the Gα_i2^βcko mice remained in the premutant state, islet DNA of Gα_i2^βcko mice showed clear recombination by appearance of an additional 390-bp band (Fig. 1, A and B). Since strong hypothalamic recombination has been reported in different Rip-Cre mouse lines (38), we analyzed DNA isolated from ctrl and Gα_i2^βcko hypothalami. The absence of the 390-bp band indicative for recombination argues against significant recombination events in the hypothalamic brain region of our mouse line (Fig. 1B). The expression pattern of Gα_i2 protein in pancreatic islets was studied by immunoblot analysis. Using specific antibodies against the COOH-terminal sequence of Gα_i2, we detected the Gα_i2 protein in islet homogenates of WT and ctrl mice, whereas Gα_i2 was strongly reduced by ∼90% in Gα_i2^βcko animals (Fig. 1, C and D), and absent in Gα_i2^−/− islet lysates (Fig. 1C). Remaining Gα_i2 immunoreactivity in Gα_i2^βcko likely results from the expression of Gα_i2 in other endocrine cell types i.e., α-, δ-, ε-, or PP-cells (see below). To further verify the cell-specific ablation of Gα_i2 in Gα_i2^βcko tissues, immunoblot analysis of hypothalamic lysates from ctrl and Gα_i2^βcko animals were performed. Gα_i2 protein was absent in hypothalamic protein extracts from Gα_i2^−/− mice (Fig. 1E), whereas the Gα_i2 protein levels of Gα_i2^βcko were unaffected, i.e., at control levels (Fig. 1, E and F). The combined analysis of Gα_i2^βcko tissues by genomic PCRs and immunoblot make it very unlikely that the studied physiological effects resulted from cells other than the pancreatic β-cells.

Fig. 1.

Gα_i2 expression analysis in β-cell-specific (Gα_i2^βcko) and globally Gα_i2-deficient (Gα_i2^−/−) mice A: schematic overview of the Gnai2 gene and PCR strategy for recombination analyses. B: genomic DNA analysis of heart, islets, and hypothalamus isolated from control (ctrl) and Gα_i2^βcko animals. Recombination is restricted to Gα_i2^βcko islets. C: representative immunoblots of islet homogenates isolated from ctrl, Gα_i2^βcko, wild-type (wt), and Gα_i2^−/− mice. Gα_i2 protein is almost absent in Gα_i2^βcko islets and completely absent in Gα_i2^−/− islets. The weak top band (*) may have resulted from Gα_i1 expression in islets. Equal loading was confirmed by β-actin detection. D: statistical analysis of Gα_i2 expression levels in ctrl and Gα_i2^βcko islets. Islets from 4 animals were analyzed in ≥3 independent experiments. E: representative immunoblot analysis of hypothalamic homogenates isolated from ctrl and Gα_i2^βcko mice. As a negative control, hypothalamic homogenates from Gα_i2^−/− mice were loaded. The top band (*) might result from Gα_i1 expression in the brain. Equal loading was confirmed by β-actin detection. F: statistical analysis of Gα_i2 expression levels in hypothalami of ctrl and Gα_i2^βcko mice. Hypothalamic tissue from 4 animals was analyzed in ≥3 independent experiments. G: representative immunostainings of pancreatic cryosections stained against Gα_i2 (yellow) and insulin (blue). Gα_i2 shows proximity to the plasma membrane, whereas insulin is mainly found in the cytoplasm. In ctrl islets, Gα_i2 staining is detectable in both insulin-positive β-cells and insulin-negative cells. Gα_i2 protein is absent in Gα_i2^βcko β-cells but can be observed in insulin-negative cells. Gα_i2 staining is completely absent in all Gα_i2^−/− islet cells. Diffuse yellow staining is background staining. Scale bars, 20 μm.

Examination of Rip-Cre-mediated recombination was performed by staining of Gα_i2 protein in pancreatic cryosections on the cellular level. β-Cells were visualized by insulin-specific antibodies (Fig. 1G). A diffuse yellow background staining was detectable; however, the specific Gα_i2 staining was restricted in proximity to the plasma membrane. Insulin (blue) was found mainly in the cytoplasm in ctrl, Gα_i2^βcko, and Gα_i2^−/− islets (left). In ctrl islets, Gα_i2 staining mainly localized to the plasma membrane (top middle) and was detectable in insulin-positive and insulin-negative cells (top right). In contrast, only a few primarily outer cells of the Gα_i2^βcko islets showed the specific membrane-close Gα_i2 staining (middle, arrows). These Gα_i2-positive cells were all identified as insulin negative and therefore non-β-cells (middle right). In Gα_i2^−/− pancreata, the membrane-close Gα_i2 staining was absent in all islet cells, insulin-negative and insulin-positive ones (bottom middle and bottom right). These results validated that Gα_i2 and insulin could be found only in ctrl β-cells, whereas Gα_i2^βcko β-cells lacked the Gα_i2 protein. Therefore, a successful recombination in Gα_i2^βcko islets was restricted to β-cells and demonstrated on a cellular level.

Analysis of G protein subunit levels in Gα_i2^−/− and Gα_i2^βcko islets of Langerhans.

Compensatory upregulation of closely related proteins is a threat of many studies using gene-deficient animals. The expression of almost all Gα_i/o isoforms in pancreatic islets as well as Gα_q, and Gα_s has been described. All of them perform major functions in the insulin secretion machinery. To get insight into possible upregulation and compensation in Gα_i2 deficiency by other G protein family members, Gα_i2^−/− (Fig. 2) and Gα_i2^βcko (Fig. 3) islet homogenates were immunoblotted for putative changes in the expression of Gα_i3, Gα_o1, Gα_o2, Gα_q, Gα_s(short), and Gα_s(long). The expression levels of the Gα_o1, Gα_o2, Gα_q, and Gα_s isoforms remained at the control values in Gα_i2^−/− (Fig. 2, A–C) and Gα_i2^βcko islets (Fig. 3, A–C). In contrast, the highly homologous isoform Gα_i3 was significantly upregulated in Gα_i2^−/− islets (Fig. 2D) (146.3 ± 14.3% of wt, P < 0.001). Importantly, in Gα_i2^βcko islets, Gα_i3 protein levels remained unchanged (119.7 ± 12.2% of control, P = 0.22; Fig. 3D).

Fig. 2.

Statistical analysis of different Gα- and Gβ-subunit expression patterns in Gα_i2^−/− islets. Gα_o1, Gα_o2 (A), Gα_s(long), Gα_s(short) (B), and Gα_q (C) expression levels are not affected by Gα_i2 deletion in Gα_i2^−/− islets. D: Gα_i3 protein upregulation is detectable in Gα_i2^−/− islets. To verify Gα_i3 antibody specificity, Gα_i3^−/− islets were loaded. E: Gβ₁ and Gβ₂ protein levels are downregulated in Gα_i2^−/− islets. Gβ₂ downregulation gains statistical significance. Insets: representative immunoblots of islet homogenates from wt and Gα_i2^−/−. Equal loading was confirmed by β-actin or α-tubulin detection.

Fig. 3.

Statistical analysis of different Gα- and Gβ-subunit expression patterns in Gα_i2^βcko islets. Gα_o1, Gα_o2 (A), Gα_s(long), Gα_s(short) (B), Gα_q (C), and Gα_i3 (D) protein expression levels are not influenced by Gα_i2-deletion in Gα_i2^βcko islets. Gα_i3 antibody specificity was shown by loading of Gα_i3^−/− islets. E: downregulation of Gβ₁ and Gβ₂ expression levels in Gα_i2^βcko islets. Gβ₂ downregulation gains statistical significance. Insets: representative immunoblots of islet homogenates from wt and Gα_i3^−/− or ctrl and Gα_i2^βcko animals. Equal loading was confirmed by β-actin detection.

G proteins are heterotrimeric proteins consisting of an α-subunit and a βγ-complex. Upon ligand activation of a GPCR, the α-subunit dissociates from the receptor and from the βγ-complex. Both the α-subunit and the βγ-complex modulate the activity of a variety of effectors, including ion channels and enzymes. Due to high Gα_i expression levels, the receptor-dependent activation of Gα_i results in the release of significant amounts of βγ-complexes (4). Moreover, it has been shown that stabilization of GDP-bound Gα by Gβγ is likely a major mechanism for the maintenance of stoichiometry between Gα and Gβγ. (9, 18). Therefore, an important point was to investigate whether expression levels of the β-isoforms were altered in Gα_i2^−/− and Gα_i2^βcko islets. Unaltered Gβ-protein levels, but simultaneously reduced Gα protein levels, might result in permanently free βγ-complexes and increased βγ-mediated effects. This could mask the physiological consequences of Gα_i2 deletion. At least five different Gβ subunits (Gβ₁–Gβ₅) have been described (13), with β₁ and β₂ being predominantly and ubiquitously expressed. In both mutant mouse lines, lower expression of Gβ₁ (Gα_i2^−/− 64.3 ± 9.6% of wt, P = 0.06; Gα_i2^βcko 61.1 ± 19.5% of ctrl, P = 0.1) and Gβ₂ (Gα_i2^−/− 80.3 ± 4.3% of wt, P < 0.01; Gα_i2^βcko 58.2 ± 8.2% of ctrl, P < 0.01) were noted in islet homogenates compared with the respective controls (Figs. 2E and 3E). However, the reduction of Gβ₂ levels was more prominent in Gα_i2^βcko islets than in Gα_i2^−/− islets, most likely due to increased Gα_i3 protein levels (145%) in Gα_i2^−/− islets. The reduced Gβ protein levels let it seem rather unlikely that effects observed in the gene-targeted mice resulted from increased Gβγ complex activity.

Metabolic evaluation of Gα_i2^−/− and Gα_i2^βcko mice.

To study the consequences of Gα_i2 deletion on the metabolism, we performed metabolic analysis of WT vs. Gα_i2^−/− (Fig. 4) and ctrl vs. Gα_i2^βcko (Fig. 5) mice. Global Gα_i2 deficiency is associated with a reduced body weight gain (22, 29), which was still observed when animals were kept under specific pathogen-free conditions (SPF) in IVCs (Fig. 4A). Differences in ad libitum-fed blood glucose levels were not obvious (Fig. 4B). Next an ip glucose tolerance test after overnight fasting was performed to analyze the impact of Gα_i2 deletion on the glucose homeostasis of these animals. Global deletion of Gα_i2 did not affect blood glucose levels during the glucose challenge (Fig. 4C). Interestingly, significantly lower insulin levels were measured in Gα_i2^−/− mice at 15, 30, and 60 min after glucose injection compared with WT mice (Fig. 4D). In addition, Gα_i2^−/− animals displayed an improvement in insulin tolerance (Fig. 4E). To assess whether β-cell-specific deletion of Gα_i2 influences the body weight of Gα_i2^βcko in a similar way as observed in Gα_i2^−/− animals, we recorded body weight and fasted blood glucose levels over time in Gα_i2^βcko and ctrl animals (Fig. 5, A and B). Continuous measurements did not reveal any differences between these two groups. Again, we monitored blood glucose levels following an ip injection of glucose after an overnight fasting period in the Gα_i2^βcko and ctrl mice (Fig. 5C). β-Cell-specific Gα_i2 deletion caused significantly increased blood glucose levels in the Gα_i2^βcko mice 30, 60, and 120 min after the glucose challenge, pointing to an impaired glucose tolerance. The parallel analysis of plasma insulin levels during the glucose tolerance test revealed significantly lower plasma insulin levels at 15, 30, and 60 min in Gα_i2^βcko mice vs. the littermate controls (Fig. 5D). Consistent with the reduced plasma insulin, plasma C-peptide levels were significantly decreased at 15 and 30 min after the glucose bolus in Gα_i2^βcko mice (Fig. 5E). Insulin sensitivity, tested by the injection of insulin to 4-h-fasted Gα_i2^βcko and ctrl mice, revealed no differences between the two groups (Fig. 5F). These findings indicate that Gα_i2 is involved in the expression, turnover, or release of insulin from pancreatic islets. Higher plasma insulin levels would have been expected from previous PTx-based studies. Surprisingly, Gα_i2 deletion restricted to β-cells lowered plasma insulin levels. Collectively, these results indicate that deletion of Gα_i2 in β-cells impairs glucose tolerance by decreasing insulin secretion, suggesting that Gα_i2 is a positive regulator of insulin secretion.

Fig. 4.

Metabolic phenotyping of Gα_i2^−/− animals. A: body weight gain of wt and Gα_i2^−/− mice. Body weight of Gα_i2^−/− mice is significantly reduced. B: blood glucose levels determined in wt and Gα_i2^−/− mice over time, which are comparable between the 2 genotypes (n = 16 animals per genotype). C: blood glucose levels in 8- to 10-wk-old wt (■) and Gα_i2^−/− () mice following ip application of glucose after overnight fasting are equal (n = 8 animals per group). D: plasma insulin levels during ip glucose tolerance testing in wt and Gα_i2^−/− animals (n = 8 animals per group). Plasma insulin levels are significantly reduced at 15, 30, and 60 min. E: insulin tolerance testing in 4-h-fasted wt and Gα_i2^−/− mice (n = 6 per genotype). *P < 0.05.

Fig. 5.

Metabolic phenotyping of Gα_i2^βcko mice. A: body weight determined in ctrl and Gα_i2^βcko mice over time. Body weight gain was similar in ctrl and Gα_i2^βcko mice (n = 8 animals per genotype). B: fasted blood glucose levels monitored in ctrl (□) and Gα_i2^βcko () mice over time. C: blood glucose levels of 8- to 10-wk-old ctrl and Gα_i2^βcko mice following ip application of glucose after overnight fasting (n = 16 animals per group). Gα_i2^βcko mice have impaired glucose tolerance. Plasma insulin (D) and plasma C-peptide (E) levels during ip glucose tolerance testing were significantly reduced in Gα_i2^βcko mice (n = 5–8 animals per group). F: insulin tolerance test in 4-h-fasted ctrl and Gα_i2^βcko mice (n = 8 animals per group). *P < 0.05, **P < 0.01, ***P < 0.001.

Islet function in vitro.

In Gα_i2^βcko mice, Gα_i2 deletion is restricted to β-cells, Gα_i3 is not upregulated in this conditional gene-targeted mouse line, and body weight is not affected by β-cell Gα_i2 deletion. Thus, these mice are an appropriate model for studying the molecular mechanism of Gα_i2 signaling in isolated islets independently of systemic effects. No abnormalities were detected in morphometric analysis of Gα_i2^βcko islets. Islet area, number of nuclei and islet size (Fig. 6, A–E), pancreas and islet insulin content, and whole islet DNA content were indistinguishable between ctrl and Gα_i2^βcko islets (Fig. 6, F–H). We performed insulin secretion studies on islets isolated from Gα_i2^βcko and ctrl mice. Insulin release was studied either under low (3 mM) or high (16 mM) glucose concentrations. All Gα_i2^βcko and ctrl islets produced a robust insulin-secretory response to high glucose concentrations, whereas insulin release was almost absent at low glucose concentrations (Fig. 7A and Table 1). Insulin secretion evoked by high K⁺ concentration, which bypasses glucose metabolism, was also not affected by the deletion of Gα_i2 in Gα_i2^βcko islets (Fig. 7A). Since a large body of PTx-based studies suggested that Gα_i2 proteins inhibit insulin secretion (11, 15, 28), we tested clonidine- and somatostatin (SST)-induced inhibition of insulin secretion. The inhibiting effects of clonidine (1 μM) and SST (100 nM) were unaffected by the absence of Gα_i2 in Gα_i2^βcko islets (Figs. 7, B and C). This argues against an inhibitory role of Gα_i2 in the regulation of insulin secretion from β-cells. In contrast, significantly reduced in vivo plasma insulin levels in Gα_i2^βcko mice suggest that Gα_i2 modulates glucose-induced insulin release by an unknown parameter, for instance, a factor present in the serum, i.e., fatty acids and/or amino acids, which signals via islet GPCRs (37). To identify the stimulus, we analyzed the effect of the free fatty acid palmitate on insulin secretion in Gα_i2^βcko islets. Simultaneous stimulation of insulin with high glucose and palmitate (200 μM) resulted in increased insulin secretion in ctrl and Gα_i2^βcko islets (Fig. 8A). Next, we studied amino acids. Application of l-leucine (100 μM) did not alter the insulin response either in ctrl or Gα_i2^βcko islets (Fig. 8B). However, application of l-arginine and l-ornithine surprisingly produced significantly differing insulin secretion responses in ctrl and Gα_i2^βcko islets. Costimulation of l-arginine (10 mM) together with glucose (16 mM) showed decreased insulin release from Gα_i2^βcko islets compared with ctrl islets (Fig. 8C). To test whether l-arginine signals via nitric oxide (NO), we used l-ornithine. In combination with glucose, l-ornithine (10 mM) was able to mimic the effect observed with l-arginine (Fig. 8D). None of the tested inhibitors or stimulators used within this setup was able to induce significant insulin secretion from isolated islets in the presence of 3 mM glucose (Table 1). To test whether our findings of significantly reduced l-arginine- and l-ornithine-induced insulin secretion from Gα_i2^βcko islets are of any physiological relevance, we performed an in vivo l-ornithine challenge. l-Ornithine was used instead of the more potent stimulus l-arginine to exclude interferences by NO-driven pathways. We studied whether a single dose of l-ornithine (1.32 mg/ml l-ornithine ip) affects plasma insulin levels in vivo in male mice. Whereas plasma insulin levels of control mice immediately increased after the l-ornithine application, plasma insulin levels of Gα_i2^βcko animals remained at basal levels (Fig. 8E). Taken together, the reduced l-ornithine-induced insulin secretion in vitro and in vivo strengthen the assumption that the PTx-sensitive Gα_i2 isoform is not an inhibitor but a stimulator of insulin secretion.

Fig. 6.

Morphometric analyses of Gα_i2^βcko islets. A: representative H&E-stained ctrl and Gα_i2^βcko islets. B: representative sections of pancreata from ctrl and Gα_i2^βcko mice visualized by immunofluorescence with anti-insulin (red) and anti-glucagon (green) antibodies. Scale bar, 40 μm. Islet size (C), number of nuclei per islet (D), and mean cell size (E) of ctrl and Gα_i2^βcko islets were equal. Morphometric analyses were performed in 6-wk-old male mice. In total, 381 control and 455 Gα_i2^βcko islets derived from 3 mice per genotype were quantified for this evaluation. Whole pancreas insulin content (F), islet insulin (G), and islet DNA content (H) of ctrl and Gα_i2^βcko mice did not differ between genotypes. Islet insulin or islet DNA content was determined following static incubation experiments (n, number of triplicates).

Fig. 7.

Insulin secretion experiments in ctrl and Gα_i2^βcko islets. A: glucose-induced insulin secretion measured at 3 and 16 mM glucose and 30 mM KCl (3 mM glucose: ctrl n = 60; Gα_i2^βcko n = 59; n, number of incubations). Inhibitory effect of clonidine (B) and SST (C) on glucose-induced insulin secretion (1 μM clonidine: ctrl n = 15; Gα_i2^βcko n = 14) was not affected by deletion of Gα_i2 in β-cells. All conditions were performed in islet preparations derived from ≥3 animals per genotype.

Table 1.

Insulin secretion evoked by 3 mM glucose concentration and in the presence of the respective substances

		Ctrl			Gαi2βcko
		Secreted insulin			Secreted insulin
Condition		%Insulin content	SE	n	%Insulin content	SE	n	P
3 mM glucose		0.189	± 0.02	28	0.174	± 0.02	27	n.s.
3 mM glucose	1 μM clonidine	0.153	± 0.05	5	0.064	± 0.01	4	n.s.
	100 nM SST	0.176	± 0.05	4	0.141	± 0.04	3	n.s.
	200 mM palmitate	0.324	± 0.04	2	0.470	± 0.02	2	n.s.
	100 μM l-leucine	0.292	± 0.05	3	0.329	± 0.06	3	n.s.
	10 mM l-arginine	0.075	± 0.02	3	0.143	± 0.03	4	n.s.
	10 mM l-ornithine	0.150	± 0.02	6	0.039	± 0.01	5	n.s.

n.s., not significant; n, set of triplicates.

Fig. 8.

Effect of different stimuli on insulin secretion experiments in ctrl and Gα_i2^βcko islets. Stimulatory effect of palmitic acid (A), and l-leucine (B) was not affected by deletion of Gα_i2 in β-cells. Potentiating effect of l-arginine (C) and l-ornithine (D) on insulin secretion was significantly reduced in Gα_i2^βcko (n, number of incubations). All conditions were performed in islet preparations derived from ≥3 animals per genotype. E: plasma insulin levels before and 1 min after ip injection of l-ornithine.

[Ca²⁺]_i measurements in islets.

To gain detailed insight into the signaling pathway, we aimed to link our in vitro findings to an intracellular correlate. It is commonly accepted that insulin secretion is a result of an increase in [Ca²⁺]_i. This can be determined in single islet cells and is therefore a cell-based correlate for β-cell insulin secretion. Measurements of intracellular calcium changes link the Gα_i-dependent signaling to insulin secretion at a cellular level. Changes in [Ca²⁺]_i signals for individual β-cells were recorded in Fura 2-loaded single islet cells. For basal conditions, β-cells were kept at 5 mM glucose before treatment with 16 mM glucose. Control and Gα_i2^βcko single islet cells showed a similar increase in [Ca²⁺]_i in response to the application of 16 mM glucose (Fig. 9A). The Ca²⁺ response was significantly reduced in Gα_i2^βcko islet cells when stimulated with l-arginine (10 mM) in the presence of basal glucose levels (5 mM) (Fig. 9B). Being aware that a simple depolarization event could be caused by the positive charge of the basic amino acid, we tested d-arginine (10 mM), the enantiomer of l-arginine. In contrast to l-arginine, d-arginine did not induce a calcium response either in Gα_i2^βcko or in ctrl single islet cells (Fig. 9C). In contrast, the l-arginine-induced response was reduced by 25% (Fig. 9E, left). The reduced response to l-arginine of Gα_i2^βcko cells was even more evident when β-cells were coexposed to high glucose concentrations and l-arginine (Fig. 9D). The l-arginine-induced increase in [Ca²⁺]_i was calculated as the difference between the maxima obtained with glucose plus l-arginine (Fig. 9D) minus the maxima obtained with glucose alone (Fig. 9A). The cotreatment resulted in a robust increase in [Ca²⁺]_i in control islets, whereas the response of Gα_i2^βcko islet cells was significantly reduced by 44% (Fig. 9E, right). The Ca²⁺ responses evoked by high K⁺ were similar in ctrl and Gα_i2^βcko β-cells (Fig. 9, D and F). These analyses identify β-cell Gα_i2 as necessary for the stimulation of insulin secretion by l-arginine in a calcium-dependent manner but independent of a depolarization that may be caused by the positive charge of the amino acid.

Fig. 9.

Calcium measurements in dispersed ctrl and Gα_i2^βcko β-cells. Representative [Ca²⁺]_i ratios (ctrl, black line; Gα_i2^βcko, gray line) elicited by 16 mM glucose (A), 10 mM l-arginine at 5 mM glucose (B), or 10 mM d-arginine at 5 mM glucose (C) (A, ctrl n = 3 mice, 465 cells; Gα_i2^βcko n = 4 mice, 298 cells; B, ctrl n = 7 mice, 606 cells; Gα_i2^βcko n = 5 mice, 687 cells; C, n = 3 mice, 447 cells; Gα_i2^βcko n = 3 mice, 413 cells). D: representative [Ca²⁺]_i response to 16 mM glucose in combination with 10 mM l-arginine and final application of KCl (ctrl n = 3 mice, 585 cells; Gα_i2^βcko n = 4 mice, 687 cells). Dotted line indicates maximum response to 16 mM glucose shown in A. E: maximum response to l-arginine was significantly reduced by 25% at 5 mM and 44% at 16 mM glucose concentrations in Gα_i2^βcko β-cells compared with ctrl β-cells, whereas the response is the difference between the maxima obtained with glucose plus l-arginine (D) minus the maxima obtained with glucose alone (A); n = 3–4 animals per group. F: maximum depolarization evoked by 30 mM KCl (n = 3–4 animals per group). All recordings were started at 5 mM glucose with additions as shown on the figure.

DISCUSSION

The present study analyzed the physiological role of Gα_i2 in insulin-secreting β-cells by using global and β-cell-specific Gα_i2 mutants. Significantly lower plasma insulin levels were observed upon glucose challenge in both the Gα_i2^βcko and the Gα_i2^−/− animal models. A normal glucose-induced insulin secretion in vitro per se accompanied these in vivo findings. Importantly, the selective and specific stimulatory effects of amino acids such as l-arginine and l-ornithine on insulin secretion were significantly diminished in Gα_i2-deficient β-cells. This was not due to alterations in islet morphology, since islet size, the number of nuclei per islet, and islet cell size were unaffected by the Gα_i2 deficiency. This study provides causal evidence for the involvement of Gα_i2 in the positive regulation of plasma insulin levels and glucose tolerance in vivo and islet insulin secretion in vitro. Although PTx-based studies in mice and isolated islets assumed an inhibitory role for all Gα_i and Gα_o proteins, our data surprisingly indicate that Gα_i2 proteins rather stimulate than inhibit the secretion of insulin by regulating l-arginine- and l-ornithine-induced insulin secretion in vitro and in vivo.

Currently, the upstream regulator of Gα_i2 is unknown. Two recent studies describe mouse models lacking the nutrient receptor GPCR family C group 6 member A (GPRC6A), which is a receptor for l-amino acids (35), in particular those with basic side chains (24, 36). Whereas Smajilovic et al. (31) found evidence for the expression of GPRC6A in pancreatic islets but found no evidence that the l-arginine-induced insulin secretion was affected by GPRC6A deletion, Pi et al. (25, 26) reported that GPRC6A mediates osteocalcin- and l-arginine-induced insulin secretion from β-cells in global and β-cell-specific GPRC6A-deficient mouse models. Our data from global and β-cell-specific Gα_i2-deficient mice showing significantly decreased plasma insulin levels following glucose challenge are consistent with those from global and β-cell-deficient GPRC6A mice (25). Future studies will have to clarify whether the concordant findings in the GPRC6A-deficient and Gα_i2-deficient mice are mechanistically connected.

Gα_i2^βcko islets secreted significantly less insulin upon costimulation with glucose and l-ornithine or l-arginine. These findings are complementary to recent studies analyzing global and β-cell-specific gene-targeted Gα_o mice (42). Those authors suggest that Gα_o is a strong inhibitor of insulin secretion, since its ablation produced a strong stimulatory effect on insulin release. Gα_o has been implicated in regulating insulin vesicle docking on β-cell membranes (42) and in inhibiting insulin release via signaling through galanin- and SST-activated GPCRs (32, 34). In line with these findings, herein we find significant inhibition of insulin secretion by SST in both control and Gα_i2^βcko islets. Of course, we are aware of the fact that these findings conflict with the expected functions for Gα_i2 based on PTx experiments, a pan inhibitor of Gα_i and Gα_o proteins and therefore a strong stimulator of insulin secretion. However, clonidine, an α₂-adrenoreceptor agonist, completely inhibited insulin release in control and Gα_i2^βcko islets. Thus, our data set provides additional evidence that Gα_i2 is not involved in the α₂-adrenoreceptor-mediated signaling. The inhibition of the inhibitory Gα_o-driven pathway by PTx may be the major cause of PTx-induced insulin release from islets.

Moreover, the PTx effect was negligible in β-cell-specific Gα_o-deficient animals (42). This suggests that, despite the fact that Gα_i proteins being expressed in islet cells and ADP-ribosylated by PTX, Gα_o is the major target for PTx stimulating insulin secretion, thereby making an inhibitory role of Gα_i2 very unlikely (42). A single study shows an inhibitory role for Gα_i2 in ghrelin-dependent inhibition of insulin secretion based on an antisense oligonucleotide approach (5). Our data demonstrate that deletion of Gα_i2 resulted in normal glucose-induced insulin secretion per se, but stimulatory effects of amino acids like l-arginine and l-ornithine on insulin secretion were diminished in Gα_i2^βcko islets. Moreover, Gα_i2^βcko mice exhibit impaired glucose tolerance caused by significantly decreased plasma insulin levels. This is also in contrast to data obtained from PTx-based studies, strengthening again the fact that Gα_i2 is an important stimulator of insulin secretion. This view is also supported by the failure of in vivo-applied l-ornithine to provoke a rise in insulin plasma levels in β-cell-specific Gα_i2-deficient animals, which was seen in littermate controls.

Our detailed analysis of G protein levels in the Gα_i2 gene-targeted mouse lines corroborates significant upregulation of Gα_i3 (about 45%) in the global Gα_i2-deficient mice, which has also been described for other tissues and organs (6, 17, 39, 40). In contrast to the global Gα_i2-deficient islets, the Gα_i3 protein levels were unchanged in Gα_i2^βcko islets. Since plasma insulin levels are reduced in both gene-targeted mouse lines, Gα_i3 is obviously not able to compensate for the lack of Gα_i2. This confirms previous reports on specific and distinct functions of Gα_i2 and Gα_i3 (8, 10, 17). Reduced Gβ expression levels parallel deletion of Gα_i2. Heterotrimeric G proteins are stoichiometrically composed of Gα and Gβγ, which dissociate selectively during the activation cycle. The stabilization of Gα and Gβγ in the trimeric complex is likely a major mechanism for the maintenance of the stoichiometric equivalence between Gα and Gβγ in the cell (9, 18). We found Gβ levels significantly reduced in both global and β-cell-specific Gα_i2-deficient mouse lines, with a more pronounced reduction of Gβ levels in Gα_i2^βcko islets (see Fig. 3E) than in Gα_i2^−/− islets (see Fig. 2E). It is reasonable to assume that the differences in the extent of reduction between Gα_i2^−/− and Gα_i2^βcko islets result from the selective upregulation of Gα_i3 protein in Gα_i2^−/− animals only, as Gα_i3 also requires Gβ as counterpart for the stabilization. This is in line with results from approaches using shRNA to downregulate Gβ isoforms in HeLa cells, which was accompanied with parallel reduced Gα protein levels in this cell line. Vice versa, several studies show that shRNA against Gα_i2 resulted in decreased Gβ₁, Gβ₂, and Gβ₄ protein levels (18, 39, 40). It has to be noted that Gβ₂ is the predominant isoform in islets, which has been reported only for human placenta (23). Signaling by the Gα_i family of “inhibitory” proteins is complex. Gα_i proteins can activate or inhibit many other effectors through either the GTP-bound Gα_i or the released Gβγ dimer upon activation. Therefore, it is important to note that both the abolished signaling of Gα_i2 via direct interaction partners and the signaling via Gβγ complexes may equally contribute to the observed phenotypes of our gene-targeted mice.

The reduced insulin levels during glucose tolerance tests were observed in global and β-cell-specific Gα_i2-deficient mice. Differences in glucose were visible only in Gα_i2^βcko animals, not in mice globally lacking Gα_i2. Wang et al. (34) also analyzed blood glucose levels of global Gα_i2-deficient mice, which were, similar to our studies, unaltered upon glucose challenge. However, those authors did not study β-cell-specific Gα_i2 mutants, and they did not report the plasma insulin levels during the glucose challenge. A reduced body weight of Gα_i2^−/− mice has been reported (29) and is also observed in our mouse line. A lean phenotype improves glucose tolerance per se, which might explain similar blood glucose levels but simultaneously decreased plasma insulin levels in global Gα_i2 mutants. In line with this, Gα_i2-deficient mice showed improved insulin tolerance, arguing for the improved peripheral insulin sensitivity. The lean phenotype was not observed in β-cell-specific Gα_i2 mutants. Control and mutant animals showed similar weight gain. However, Gα_i2^βcko animals showed impaired glucose tolerance caused by impairment of the β-cells' ability to secrete the required amount of insulin. The differences in blood glucose levels between global and β-cell-specific Gα_i2 mice imply that the metabolic functions of Gα_i2 in the body are broader than previously thought. Gα_i2 signaling might play a role for glucose homeostasis in liver, white adipose tissue, or skeletal muscle as well, but Gα_i2 signaling is also important for efficient pancreatic β-cell function. Thus, we propose that the diminished plasma insulin levels indeed result from islet Gα_i2 dysfunction, caused by an impaired potentiating effect of l-arginine and l-ornithine in pancreatic β-cells. In addition, the reduced increases of intracellular calcium levels upon l-arginine treatment in Gα_i2^βcko β-cells strengthen the finding that Gα_i2 stimulates insulin secretion in a calcium-dependent manner. However, it remains elusive whether regulation of intracellular calcium release and/or extracellular calcium influx is disturbed by the Gα_i2 deletion.

In conclusion, the experiments presented identify Gα_i2 as an endogenous positive modulator of the l-arginine- and l-ornithine-induced stimulation of insulin secretion, which may be an important feedback mechanism for fine-tuning amino acid plasma levels and insulin secretion. Interestingly, the PTx-sensitive Gα_i2 isoform was identified as a stimulator of insulin secretion. This strengthens the hypothesis that Gα_o and Gα_i proteins have distinct, nonredundant, and importantly opposite roles in pancreatic β-cells: 1) Gα_o isoforms function as inhibitors of insulin secretion and are the main isoform being responsible for the dysregulated insulin release upon PTx treatment; and 2) Gα_i2 isoforms stimulate insulin secretion upon l-arginine and l-ornithine treatment.

GRANTS

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) with grants to B. Nürnberg and V. Leiss, and GRK 1302 to B. Nürnberg, by the Intramural Program of the National Institutes of Health to L. Birnbaumer (Project Z01 ES-101684), and by the German Ministry of Education and Research (BMBF: DZD, 01GI0922) to A. Schürmann.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: V.L., A. Schürmann, C.H., and B.N. conception and design of research; V.L., K.F., A.N., M.R., and A. Schönsiegel performed experiments; V.L., K.F., M.R., and A. Schönsiegel analyzed data; V.L., K.F., L.B., A. Schürmann, C.H., and B.N. interpreted results of experiments; V.L. and K.F. prepared figures; V.L., C.H., and B.N. drafted manuscript; V.L., K.F., A.N., L.B., A. Schürmann, C.H., and B.N. edited and revised manuscript; V.L., K.F., A.N., M.R., A. Schönsiegel, L.B., A. Schürmann, C.H., and B.N. approved final version of manuscript.