*Article* **The Capacity to Secrete Insulin Is Dose-Dependent to Extremely High Glucose Concentrations: A Key Role for Adenylyl Cyclase**

**Katherine M. Gerber 1,†, Nicholas B. Whitticar 2,3,†, Daniel R. Rochester <sup>2</sup> , Kathryn L. Corbin <sup>2</sup> , William J. Koch 2,3 and Craig S. Nunemaker 2,4,\***



**Citation:** Gerber, K.M.; Whitticar, N.B.; Rochester, D.R.; Corbin, K.L.; Koch, W.J.; Nunemaker, C.S. The Capacity to Secrete Insulin Is Dose-Dependent to Extremely High Glucose Concentrations: A Key Role for Adenylyl Cyclase. *Metabolites* **2021**, *11*, 401. https://doi.org/ 10.3390/metabo11060401

Academic Editors: Belinda Yau and Melkam Kebede

Received: 31 May 2021 Accepted: 18 June 2021 Published: 19 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Abstract:** Insulin secretion is widely thought to be maximally stimulated in glucose concentrations of 16.7-to-30 mM (300-to-540 mg/dL). However, insulin secretion is seldom tested in hyperglycemia exceeding these levels despite the Guinness World Record being 147.6 mM (2656 mg/dL). We investigated how islets respond to 1-h exposure to glucose approaching this record. Insulin secretion from human islets at 12 mM glucose intervals dose-dependently increased until at least 72 mM glucose. Murine islets in 84 mM glucose secreted nearly double the insulin as in 24 mM (*p* < 0.001). Intracellular calcium was maximally stimulated in 24 mM glucose despite a further doubling of insulin secretion in higher glucose, implying that insulin secretion above 24 mM occurs through amplifying pathway(s). Increased osmolarity of 425-mOsm had no effect on insulin secretion (1-h exposure) or viability (48-h exposure) in murine islets. Murine islets in 24 mM glucose treated with a glucokinase activator secreted as much insulin as islets in 84 mM glucose, indicating that glycolytic capacity exists above 24 mM. Using an incretin mimetic and an adenylyl cyclase activator in 24 mM glucose enhanced insulin secretion above that observed in 84 mM glucose while adenylyl cyclase inhibitor reduced stimulatory effects. These results highlight the underestimated ability of islets to secrete insulin proportionally to extreme hyperglycemia through adenylyl cyclase activity.

**Keywords:** amplifying pathway; hyperglycemia; adenylyl cyclase; incretins; glucokinase; forskolin; cAMP; exenatide; diabetes; insulin; islets

## **1. Introduction**

Pancreatic beta cells secrete insulin in response to glucose stimulation to maintain blood glucose levels within a relatively narrow range [1]. Insulin is required to transport glucose from the bloodstream to target tissues. High blood glucose levels in the body are caused by problems with insulin secretion, insulin action, or both. Extremely high levels of glucose lead to the presentation of ketoacidosis or hyperosmolar hyperglycemic nonketoic syndrome, which are key indicators of the metabolic disease diabetes [2]. Poor control of glucose regulation in this disease can bring potential stupor, coma, or death [3].

In humans, insulin secretion is typically stimulated by glucose concentrations ranging from 4.4 to 6.6 mM (80–120 mg/dL) [2,4]. Moreover, it is generally accepted that the effective concentration of glucose for half of the maximal insulin secretion (EC50) is approximately 5 mM. These EC<sup>50</sup> estimations are based on dose-response curves with the highest stimulation typically being described at 16.7 (300 mg/dL) to 30 mM (540 mg/dL). However, close inspection of published glucose dose-response curves suggests that even

though a classic sigmoid dose-response curve should flatten out, the curves typically show an increasing trend of higher insulin secretion near the maximal glucose level tested. This provokes the hypothesis that insulin secretion may be sensitive to a much wider range of glucose concentrations than commonly thought [5–9].

Interestingly, there are multiple case reports of individuals who have had blood glucose levels greater than 100 mM (1800 mg/dL) and survived [2,4,10,11]. This includes the world record blood glucose level of 147.6 mM (2656 mg/dL) which was set by a young boy when admitted to the hospital [11]. The pathways involving insulin secretion at extremely high glucose concentrations have not been examined to our knowledge.

Glucose is the primary stimulator of insulin release from pancreatic beta cells. The ability of glucose to elicit an increase in intracellular calcium leading to insulin secretion is known as the triggering pathway [12–14]. This pathway begins when glucose enters the beta cell through both the GLUT 1 and GLUT 3 glucose transporters in human islets and the GLUT 2 transporter in mouse islets [13,14]. Glucose is then phosphorylated by glucokinase and yields glucose-6-phosphate which travels through glycolysis to yield pyruvate and ATP. Glycolysis and downstream mitochondrial metabolism drive the ratio of ATP to ADP to rise, leading to the closure of KATP channels. This closure activates the voltage-dependent calcium channels to allow the influx of calcium that constitutes the triggering pathway for insulin secretion.

In addition to the triggering pathway, many different intermediate metabolites of glucose and other cellular components are thought to participate in a series of events known as the amplification pathway(s) in which KATP channel closure is not the source of the increased insulin secretion. A review on this topic contained within this special issue describes this pathway as "the sequence of events that enables the secretory response to a nutrient secretagogue to exceed the secretory response of a purely depolarizing stimulus" [15]. Cyclic adenosine monophosphate (cAMP) is a powerful player in the amplification pathway-related insulin secretion. Glucose is a known factor leading to the upregulation of cAMP which is formed from ATP via adenylyl cyclases [16], although the exact mechanism of metabolism stimulated cAMP is unknown. Additionally, there is a glucose-linked amplification pathway that augments insulin secretion through adenylyl cyclase (AC) activation caused by incretin stimulation. Incretins are hormones secreted by endocrine cells in the small intestine after meal ingestion that lead to insulin secretion [12]. Gastric inhibitory polypeptide (GIP) and glucagon-like peptide-1 (GLP-1) are specific incretins that act on the beta cell. Incretins interact with G-protein-coupled receptors at the cell membrane and lead to an upregulation of cAMP [17,18]. The common theme is that cAMP is involved in the early steps leading to insulin secretion in these amplifying pathway(s).

In this study, we aim to determine the relationship between glucose stimulation and insulin secretion in a much higher glucose range than typically examined. We show that insulin secretion is maintained in response to extreme glucose, osmolarity does not affect insulin secretion or islet viability at extreme levels, and that intracellular calcium is maximized at 24 mM glucose. We also sought to identity possible pathway(s) used in extreme glucose conditions and found that adenylyl cyclase plays an important role.

#### **2. Results**

#### *2.1. Insulin Secretion Is Maintained in Response to Extreme Glucose*

Identifying the maximal physiological response of biological systems often gives insight into function. We wanted to determine what happens to insulin secretion above the commonly accepted maximum concentration of glucose for stimulation. We thus examined insulin secretion from human and murine islets in conditions of extremely high glucose. Islets from human donors were placed in mannitol-balanced Krebs-Ring buffer (KRB) solutions containing glucose ranging from 0 mM to 144 mM for 1 h. As shown in Figure 1A, insulin secretion from each of the four donors was normalized to their respective maximal value and averaged amongst donors (see Table 1 for donor information). From 0 to 24 mM,

there was a significant upward trend (rho = 0.93, *p* < 0.001) and from 24–84 mM, there was still a positive correlation, though not as strong (rho = 0.73, *p* < 0.001). Insulin secretion from murine islets was also measured (Figure 1B). From 0–24 mM, there was a strong upward trend (rho = 0.75, *p* < 0.001), and from 24–84 mM, there was still a significant positive trend (rho = 0.63, *p* < 0.001). Overall, there was a positive correlation between insulin and glucose for 0–24 mM, as expected, but a similar relationship exists from 24 mM glucose and beyond. tive maximal value and averaged amongst donors (see Table 1 for donor information). From 0 to 24 mM, there was a significant upward trend (rho = 0.93, *p* < 0.001) and from 24–84 mM, there was still a positive correlation, though not as strong (rho = 0.73, *p* < 0.001). Insulin secretion from murine islets was also measured (Figure 1B). From 0–24 mM, there was a strong upward trend (rho = 0.75, *p* < 0.001), and from 24–84 mM, there was still a significant positive trend (rho = 0.63, *p* < 0.001). Overall, there was a positive correlation between insulin and glucose for 0–24 mM, as expected, but a similar relationship exists from 24 mM glucose and beyond.

Figure 1A, insulin secretion from each of the four donors was normalized to their respec-

*Metabolites* **2021**, *11*, x FOR PEER REVIEW 3 of 14

**Figure 1.** Human and murine islet insulin secretion in conditions of extremely high glucose. Islets from human donors (**A**) and mice (**B**) were placed in mannitol balanced KRB solutions containing glucose ranging from 0 mM to 144 mM for 1 h. Insulin secretion from each donor or mouse replicate was normalized to their respective maximal value and averaged. Dotted linear trendlines are drawn for 0–24 mM and 24–84 mM glucose to indicate slope, Spearman's rho, and the respective *p*-value. **Figure 1.** Human and murine islet insulin secretion in conditions of extremely high glucose. Islets from human donors (**A**) and mice (**B**) were placed in mannitol balanced KRB solutions containing glucose ranging from 0 mM to 144 mM for 1 h. Insulin secretion from each donor or mouse replicate was normalized to their respective maximal value and averaged. Dotted linear trendlines are drawn for 0–24 mM and 24–84 mM glucose to indicate slope, Spearman's rho, and the respective *p*-value.



#### *2.2. Increased Osmolarity Does Not Alter Insulin Secretion or Islet Viability 2.2. Increased Osmolarity Does Not Alter Insulin Secretion or Islet Viability*

To control for osmolarity as a variable, solutions were balanced to ~425 mOsm/L using the sugar alcohol mannitol. Mannitol was used to account for the difference in osmolarity between low and high glucose conditions, and here the effect of mannitol on its own was explored. Mannitol is not metabolized and therefore should not increase intracellular calcium or insulin secretion, although indirect factors such as membrane potential or ion flux might alter insulin secretion [19]. To investigate this possibility, islets were treated for 1 h in modified KRB with 0 mM glucose or 84 mM glucose with and without mannitol balanced to 144 mM. While it is known that various osmotic receptors on beta-cells have been shown to alter insulin secretion [20], as shown in Figure 2A, large increases in osmolarity caused no significant change in insulin secretion, and mannitol alone did not have To control for osmolarity as a variable, solutions were balanced to ~425 mOsm/L using the sugar alcohol mannitol. Mannitol was used to account for the difference in osmolarity between low and high glucose conditions, and here the effect of mannitol on its own was explored. Mannitol is not metabolized and therefore should not increase intracellular calcium or insulin secretion, although indirect factors such as membrane potential or ion flux might alter insulin secretion [19]. To investigate this possibility, islets were treated for 1 h in modified KRB with 0 mM glucose or 84 mM glucose with and without mannitol balanced to 144 mM. While it is known that various osmotic receptors on beta-cells have been shown to alter insulin secretion [20], as shown in Figure 2A, large increases in osmolarity caused no significant change in insulin secretion, and mannitol alone did not have a stimulatory effect. As expected, islets in each 84 mM glucose group

*Metabolites* **2021**, *11*, x FOR PEER REVIEW 4 of 14

secreted significantly more insulin (*p* <0.01) than islets in each glucose-free group, but osmolarity changes due to mannitol did not significantly affect insulin secretion. more insulin (*p* <0.01) than islets in each glucose-free group, but osmolarity changes due to mannitol did not significantly affect insulin secretion.

a stimulatory effect. As expected, islets in each 84 mM glucose group secreted significantly

**Figure 2.** Effects of high glucose and mannitol on mouse islet insulin secretion and cell death. (**A**) Mouse islets were placed in wells containing modified KRB with 0 mM or 84 mM glucose ± mannitol balanced to 144 mM total for 1h. Increasing osmolarity using mannitol did not significantly alter insulin secretion in glucose-free or high-glucose solutions. Twenty islets per condition in duplicate ran in 2 separate trials (N = 4). (**B**) Islets were placed in either standard RPMI, high glucose (144 mM), high mannitol (144 mM), or both glucose (60mM) and mannitol (84 mM). Islets treated overnight with 5 ng/mL IL-1beta and 10 ng/mL TNF-alpha were used as a positive control to show typical fluorescence levels of induced cell death. Islets in each condition ranged from N=31–37 after combining two separate trials. All data are presented as mean ± SEM. \*\* *p* < 0.01, \*\*\* *p* < 0.001, N.S. = Not Significant. **Figure 2.** Effects of high glucose and mannitol on mouse islet insulin secretion and cell death. (**A**) Mouse islets were placed in wells containing modified KRB with 0 mM or 84 mM glucose ± mannitol balanced to 144 mM total for 1h. Increasing osmolarity using mannitol did not significantly alter insulin secretion in glucose-free or high-glucose solutions. Twenty islets per condition in duplicate ran in 2 separate trials (N = 4). (**B**) Islets were placed in either standard RPMI, high glucose (144 mM), high mannitol (144 mM), or both glucose (60mM) and mannitol (84 mM). Islets treated overnight with 5 ng/mL IL-1beta and 10 ng/mL TNF-alpha were used as a positive control to show typical fluorescence levels of induced cell death. Islets in each condition ranged from N=31–37 after combining two separate trials. All data are presented as mean ± SEM. \*\* *p* < 0.01, \*\*\* *p* < 0.001, N.S. = Not Significant.

> Next, we measured cell death using fluorescent microscopy to check for toxicity. Isolated mouse islets treated in standard RPMI, high glucose (144 mM), high osmolarity (144 mM), or a combination of both (84 mM glucose + 60 mM mannitol) for 48 h displayed no significant differences in cell death (propidium iodide) or apoptosis (annexin V) (Figure 2B). Proinflammatory cytokines were used as a positive control to induce beta-cell death. Collectively, these results indicate that substantial increases in osmolarity have no effect on cellular function or viability in mouse islets in these conditions. Next, we measured cell death using fluorescent microscopy to check for toxicity. Isolated mouse islets treated in standard RPMI, high glucose (144 mM), high osmolarity (144 mM), or a combination of both (84 mM glucose + 60 mM mannitol) for 48 h displayed no significant differences in cell death (propidium iodide) or apoptosis (annexin V) (Figure 2B). Proinflammatory cytokines were used as a positive control to induce beta-cell death. Collectively, these results indicate that substantial increases in osmolarity have no effect on cellular function or viability in mouse islets in these conditions.

#### *2.3. Intracellular Calcium Is Maximally Stimulated in 24 mM Glucose 2.3. Intracellular Calcium Is Maximally Stimulated in 24 mM Glucose*

To determine whether these responses were calcium dependent, mouse islets were loaded with the calcium probe fura-2AM and exposed to increasing stimulation as indicated by the black bars in Figure 3. As expected, a large increase was observed in response to 24 mM glucose stimulation. When the stimulus was increased from 24 to 84 mM glucose, average calcium levels of all islets were not significantly different (Figure 3A). Large increases in glucose above 24 mM do not appear to impact calcium handling. Furthermore, there was no significant difference between calcium levels in the 3 mM and the 3 mM wash (*p* > 0.05) showing that the extreme glucose did not alter the islet's ability to control the secretion process. To determine whether these responses were calcium dependent, mouse islets were loaded with the calcium probe fura-2AM and exposed to increasing stimulation as indicated by the black bars in Figure 3. As expected, a large increase was observed in response to 24 mM glucose stimulation. When the stimulus was increased from 24 to 84 mM glucose, average calcium levels of all islets were not significantly different (Figure 3A). Large increases in glucose above 24 mM do not appear to impact calcium handling. Furthermore, there was no significant difference between calcium levels in the 3 mM and the 3 mM wash (*p* > 0.05) showing that the extreme glucose did not alter the islet's ability to control the secretion process.

Tolbutamide, a sulfonylurea known to depolarize beta cells via reduction of potassium permeability, leads to the opening of voltage-dependent calcium channels to subsequently trigger insulin release [21]. As anticipated, there was no significant difference (*p* > 0.05) between 24 mM and 24 mM combined with 250 µM tolbutamide in terms of intracellular calcium influx (Figure 3B). This supports the observation that intracellular calcium is maximally stimulated in 24 mM glucose.

**A. B.**

*Metabolites* **2021**, *11*, x FOR PEER REVIEW 5 of 14

**Figure 3.** Effects of high glucose on intracellular calcium. (**A**) Intracellular calcium was measured in islets exposed to the increasing glucose conditions noted by the horizontal bars. Calcium levels increased sharply from 3 mM to 24 mM glucose but did not increase further when exposed to 84 mM glucose. Average calcium levels from 23 islets were calculated from the 15–20- and 30–35 min time points and showed no significant difference. (**B**) Intracellular calcium was measured in islets exposed to the increasing glucose conditions and tolbutamide noted by the horizontal bars. Calcium levels increased sharply from 3 mM to 24 mM but did not increase further when exposed to 24 mM glucose combined with 250 μM tolbutamide. Average calcium levels from 23 islets were calculated from the 15–20 and 30–35 min time points and showed no significant difference. \*\*\* *p* < 0.001 difference between 3 and 24 mM glucose. **Figure 3.** Effects of high glucose on intracellular calcium. (**A**) Intracellular calcium was measured in islets exposed to the increasing glucose conditions noted by the horizontal bars. Calcium levels increased sharply from 3 mM to 24 mM glucose but did not increase further when exposed to 84 mM glucose. Average calcium levels from 23 islets were calculated from the 15–20- and 30–35 min time points and showed no significant difference. (**B**) Intracellular calcium was measured in islets exposed to the increasing glucose conditions and tolbutamide noted by the horizontal bars. Calcium levels increased sharply from 3 mM to 24 mM but did not increase further when exposed to 24 mM glucose combined with 250 µM tolbutamide. Average calcium levels from 23 islets were calculated from the 15–20 and 30–35 min time points and showed no significant difference. \*\*\* *p* < 0.001 difference between 3 and 24 mM glucose. amide. Average calcium levels from 23 islets were calculated from the 15–20 and 30–35 min time points and showed no significant difference. \*\*\* *p* < 0.001 difference between 3 and 24 mM glucose. Tolbutamide, a sulfonylurea known to depolarize beta cells via reduction of potassium permeability, leads to the opening of voltage-dependent calcium channels to subsequently trigger insulin release [21]. As anticipated, there was no significant difference (*p*  > 0.05) between 24 mM and 24 mM combined with 250 µM tolbutamide in terms of intracellular calcium influx (Figure 3B). This supports the observation that intracellular calcium is maximally stimulated in 24 mM glucose.

#### sium permeability, leads to the opening of voltage-dependent calcium channels to subse-*2.4. Glycolytic Capacity to Secrete Insulin Dose-Dependently Extends above 24 mM Glucose 2.4. Glycolytic Capacity to Secrete Insulin Dose-Dependently Extends above 24 mM Glucose*

Tolbutamide, a sulfonylurea known to depolarize beta cells via reduction of potas-

quently trigger insulin release [21]. As anticipated, there was no significant difference (*p*  > 0.05) between 24 mM and 24 mM combined with 250 µM tolbutamide in terms of intracellular calcium influx (Figure 3B). This supports the observation that intracellular calcium is maximally stimulated in 24 mM glucose. *2.4. Glycolytic Capacity to Secrete Insulin Dose-Dependently Extends above 24 mM Glucose* We next conducted a series of trials to determine what drives increased insulin secretion in high glucose. The following studies were all conducted in 24 mM glucose (regarded as maximal) and 84 mM glucose (regarded as extreme). As shown in Figure 4A, insulin secretion was significantly higher in 84 mM glucose compared to 24 mM glucose after combining all trials (*p* < 0.001). The averages are reproduced for comparisons made in Figures 4B and 5. We next conducted a series of trials to determine what drives increased insulin secretion in high glucose. The following studies were all conducted in 24 mM glucose (regarded as maximal) and 84 mM glucose (regarded as extreme). As shown in Figure 4A, insulin secretion was significantly higher in 84 mM glucose compared to 24 mM glucose after combining all trials (*p* < 0.001). The averages are reproduced for comparisons made in Figure 4B and Figure 5.

We next conducted a series of trials to determine what drives increased insulin secre-

**Figure 4.** Insulin secretion in extremely high glucose is associated with increased glycolysis. (**A**) Insulin secretion was measured after 1h incubation in 24 mM glucose (regarded as maximal) and 84 mM glucose (regarded as extreme) shown **Figure 4.** Insulin secretion in extremely high glucose is associated with increased glycolysis. (**A**) Insulin secretion was measured after 1h incubation in 24 mM glucose (regarded as maximal) and 84 mM glucose (regarded as extreme) shown for four separate trials with N = 2–3 replicates to demonstrate consistency of the difference in insulin secretion from trial to trial. Combining these trials, the data were significantly different with a *p*-value of <0.001 by two-tailed T-test between **Figure 4.** Insulin secretion in extremely high glucose is associated with increased glycolysis. (**A**) Insulin secretion was measured after 1h incubation in 24 mM glucose (regarded as maximal) and 84 mM glucose (regarded as extreme) shown for four separate trials with N = 2–3 replicates to demonstrate consistency of the difference in insulin secretion from trial to trial. Combining these trials, the data were significantly different with a *p*-value of <0.001 by two-tailed T-test between 24 and 84 mM glucose. (**B**) Mouse islets were placed in wells containing modified KRB with 24 mM or 84 mM glucose ± mannitol balanced to 84 mM total for 1 h with or without 500 nM GKA. The GKA increased insulin secretion significantly in 24 mM glucose (*p* <0.01) but failed to do so in 84 mM glucose. All data are presented as ± SEM. N = 3–11 replicates. \* *p* <0.05, \*\* *p* <0.01, \*\*\* *p* <0.001.

for four separate trials with N = 2–3 replicates to demonstrate consistency of the difference in insulin secretion from trial

<0.05, \*\* *p* <0.01, \*\*\* *p* <0.001.

24 and 84 mM glucose. (**B**) Mouse islets were placed in wells containing modified KRB with 24 mM or 84 mM glucose ± mannitol balanced to 84 mM total for 1 h with or without 500 nM GKA. The GKA increased insulin secretion significantly in 24 mM glucose (*p* <0.01) but failed to do so in 84 mM glucose. All data are presented as ± SEM. N = 3–11 replicates. \* *p*

> To examine the contribution of glycolytic activity leading to insulin secretion, the glucokinase activator R0-28-1675 (hereafter called GKA) was utilized, and insulin measurements were taken. As shown in Figure 4B, islets treated with GKA in 24 mM glucose displayed a significant increase in insulin secretion compared to islets in 24 mM alone (*p*  <0.01). Insulin secretion in 84 mM glucose was on par with islets given GKA in 24 mM glucose, but GKA did not provide any additional stimulation in 84 mM glucose. Together, these data indicate that glycolytically driven insulin secretion continues to dose-dependently increase, nearly doubling insulin secretion between 24 and 84 mM glucose.

**Figure 5.** cAMP-related drug effect on insulin secretion in high glucose. (**A**–**C**) Murine islets were placed in wells containing modified KRB with 24 mM or 84 mM glucose ± mannitol balanced to 84 mM total for 1 h and then various drugs were added to each condition for 1 h of incubation before insulin was measured: 10 µM forskolin (**A**), 10 µM MDL 12,330A (**B**), and 10 nM exendin-4 (**C**) All data are presented as ± SEM. N = 5-11 replicates. \* *p* <0.05, \*\* *p* <0.01, \*\*\* *p* <0.001. **Figure 5.** cAMP-related drug effect on insulin secretion in high glucose. (**A**–**C**) Murine islets were placed in wells containing modified KRB with 24 mM or 84 mM glucose ± mannitol balanced to 84 mM total for 1 h and then various drugs were added to each condition for 1 h of incubation before insulin was measured: 10 µM forskolin (**A**), 10 µM MDL 12,330A (**B**), and 10 nM exendin-4 (**C**) All data are presented as ± SEM. N = 5–11 replicates. \* *p* <0.05, \*\* *p* <0.01, \*\*\* *p* <0.001.

*2.5. The cAMP Pathway Provides Additional Capacity for Insulin Secretion in Extremely High Glucose* cAMP, a known amplifier of insulin secretion, was examined by the use of forskolin, an adenylyl cyclase activator that increases cAMP levels. As shown in Figure 5A, in 24 mM glucose for 1 h, forskolin increased insulin secretion dramatically compared to the control 24 mM glucose (*p* < 0.001). There was also a significant difference found between 84 mM glucose and 84 mM glucose combined with forskolin (*p* < 0.001). In addition, the amount of insulin secreted at 84 mM glucose with forskolin was significantly higher than To examine the contribution of glycolytic activity leading to insulin secretion, the glucokinase activator R0-28-1675 (hereafter called GKA) was utilized, and insulin measurements were taken. As shown in Figure 4B, islets treated with GKA in 24 mM glucose displayed a significant increase in insulin secretion compared to islets in 24 mM alone (*p* < 0.01). Insulin secretion in 84 mM glucose was on par with islets given GKA in 24 mM glucose, but GKA did not provide any additional stimulation in 84 mM glucose. Together, these data indicate that glycolytically driven insulin secretion continues to dose-dependently increase, nearly doubling insulin secretion between 24 and 84 mM glucose.

#### the amount of insulin secreted at 24 mM glucose with forskolin (*p* < 0.01). Overall, the cAMP pathway has the capacity to dramatically increase insulin secretion, even at extreme glucose concentrations. *2.5. The cAMP Pathway Provides Additional Capacity for Insulin Secretion in Extremely High Glucose*

MDL-12,330A (MDL), an adenylyl cyclase inhibitor, was used to determine if reducing cAMP by inhibiting adenylyl cyclase could block the additional insulin secretion in cAMP, a known amplifier of insulin secretion, was examined by the use of forskolin, an adenylyl cyclase activator that increases cAMP levels. As shown in Figure 5A, in 24 mM glucose for 1 h, forskolin increased insulin secretion dramatically compared to the control 24 mM glucose (*p* < 0.001). There was also a significant difference found between 84 mM glucose and 84 mM glucose combined with forskolin (*p* < 0.001). In addition, the amount of insulin secreted at 84 mM glucose with forskolin was significantly higher than the amount of insulin secreted at 24 mM glucose with forskolin (*p* < 0.01). Overall, the cAMP pathway has the capacity to dramatically increase insulin secretion, even at extreme glucose concentrations.

MDL-12,330A (MDL), an adenylyl cyclase inhibitor, was used to determine if reducing cAMP by inhibiting adenylyl cyclase could block the additional insulin secretion in extreme hyperglycemic conditions. MDL appeared to partially inhibit insulin secretion in 84 mM glucose. The significantly higher levels of insulin secretion in 84 mM glucose compared to 24 mM glucose were not significant with MDL added to 84 mM glucose (Figure 5B). Importantly, MDL did not reduce insulin secretion in 24 mM glucose compared to 24 mM

glucose alone, indicating that the inhibitory effects of MDL on adenylyl cyclase occur only in extremely high glucose conditions when adenylyl cyclase is likely contributing more to overall insulin secretion.

Pancreatic islets were next stimulated with exendin-4, a GLP-1 agonist, to examine the possible contributions of incretin pathways on insulin secretion in extreme glucose conditions. There was a significant increase in insulin secretion with exendin added to 24 mM glucose compared to 24 mM glucose alone (*p* < 0.05). However, when exendin was tested with 84 mM glucose, there was no significant difference between 84 mM glucose and 84 mM glucose with exendin (Figure 5C). This indicates that insulin secretion, due to incretin effects, continues to increase between 24 mM and 84 mM glucose, revealing possible incretin involvement in the amplification pathway at extreme glucose concentrations.

#### **3. Discussion**

#### *3.1. Maximum Glucose Concentrations for Glucose-Stimulated Insulin Secretion Are Much Higher Than Previously Reported*

Our study demonstrates a surprisingly high capacity of islets to maintain stimulussecretion coupling in glucose concentrations far exceeding what is considered normal for both mice and humans. We showed that insulin secretion doubled from 24 mM to 84 mM glucose, while calcium levels were unchanged over the same range. This suggests that insulin secretion occurs through the amplifying pathway, which can contribute 50% or more to insulin secretion when calcium is saturated [22]. In addition, our findings suggest that there is still much to be learned about the mechanisms and limits of the amplifying pathway of glucose-stimulated insulin secretion.

#### *3.2. High Glucose Leads to Increased cAMP via Different Potential Mechanisms*

Insulin secretion stimulated by cAMP can be thought of in two independent routes, the secretion that is dependent on calcium changes and that which is stimulated from cellular glucose metabolism [16]. Once cAMP accumulates, intracellular calcium rises through increased L-type-calcium-channel activity [23,24] and through release from intracellular stores [25]. Although cAMP thus has the capacity to increase intracellular calcium, our results reveal that intracellular calcium is already saturated in 24 mM glucose, so any additional stimulation of insulin secretion would likely not involve changes in calciumdependent pathways. This study demonstrates that extremely high glucose levels can increase insulin secretion independently of the triggering pathway.

It is understood that increased glucose leads to an increase in cAMP [26]. Our data show that glucose continues to dose-dependently increase rates of glycolysis, as shown by a near doubling of insulin secretion between 24 and 84 mM glucose. As shown in Figure 6, the effects of extreme glucose can be reproduced with GKA in 24 mM glucose. This indicates that glycolytically driven metabolite formation is responsible for a large portion of insulin secreted above 24 mM glucose. Additionally, it should be noted that in Figure 6, forskolin and MDL act directly on adenylyl cyclase to alter cAMP levels. Forskolin, which acts to increase adenylyl cyclase, showed a huge potential for stimulating insulin secretion at high glucose concentrations. With that, MDL was able to partially block the insulin released in extremely high glucose conditions. MDL may not have fully reduced the insulin secreted in 84 mM glucose back down to 24 mM levels as expected, since other pathways independent of cAMP are likely working synergistically to amplify insulin secretion. Regardless, there are several ways in which increased glucose metabolism can augment insulin secretion through cAMP that do not require changes in intracellular calcium.

cium.

augment insulin secretion through cAMP that do not require changes in intracellular cal-

**Figure 6.** The triggering and amplifying pathways of insulin secretion in extreme glucose. (Left) Glucose enters the beta cell and is metabolized, causing an increase in the ATP to ADP ratio. ATP-sensitive potassium channels (KATP) close, causing membrane depolarization, the opening of voltage-dependent calcium channels (VDCC), and insulin secretion. (Right) GLP-1 from alpha cells or exendin stimulate beta-cell GLP-1 receptors which stimulate AC. Additionally, various metabolites produced from glucose metabolism can stimulate AC. Increased production of cAMP stimulates PKA and EPAC2A, causing increased insulin secretion by various mechanisms. Ovals indicate points of stimulation (green) or inhibition (red) for pharmacological agents used in these studies. **Figure 6.** The triggering and amplifying pathways of insulin secretion in extreme glucose. (Left) Glucose enters the beta cell and is metabolized, causing an increase in the ATP to ADP ratio. ATP-sensitive potassium channels (KATP) close, causing membrane depolarization, the opening of voltage-dependent calcium channels (VDCC), and insulin secretion. (Right) GLP-1 from alpha cells or exendin stimulate beta-cell GLP-1 receptors which stimulate AC. Additionally, various metabolites produced from glucose metabolism can stimulate AC. Increased production of cAMP stimulates PKA and EPAC2A, causing increased insulin secretion by various mechanisms. Ovals indicate points of stimulation (green) or inhibition (red) for pharmacological agents used in these studies.

been shown to be glucose dependent [26]. This should be no surprise, since cAMP/PKA pathways have been long explored as targets for potential diabetes therapy [27]. As shown in Figure 6, cAMP activates downstream effectors PKA and guanine-nucleotide exchange protein (EPAC), which both lead to cAMP-mediated insulin secretion [28]. Focusing on the PKA pathway, glutamate derived from glucose through the malate aspartate shuttle is the specific signal underlying insulin secretion after being stimulated via cAMP [12]. It has been shown that the cAMP/PKA pathway potentiates the release of insulin via increased effectiveness of KATP-channel-independent actions of glucose [29], which is consistent with our observations. Intracellular calcium is unaffected by PKA activation, and PKA effects on insulin secretion are mediated by the phosphorylation of various downstream proteins [30]. The mechanisms leading to the rise in insulin secretion are studied with electrophysiological and optical methods which monitor the movements and exocytosis of individual insulin granules [31,32]. Changes in the size of distinct granular pools, facilitation of granule recruitment from the pools to the plasma membrane, and the acceleration of the priming process that confers granules with release competence may all play a role as well as the involvement of SNARE complexes. Relating this back to our study, the PKA pathway stemming from the increase in cAMP ties to the amplification pathway via glucose metabolism and is most likely a key player in insulin release at such high glucose concentrations. Protein kinase A (PKA) has a permissive role in increasing insulin secretion that has been shown to be glucose dependent [26]. This should be no surprise, since cAMP/PKA pathways have been long explored as targets for potential diabetes therapy [27]. As shown in Figure 6, cAMP activates downstream effectors PKA and guanine-nucleotide exchange protein (EPAC), which both lead to cAMP-mediated insulin secretion [28]. Focusing on the PKA pathway, glutamate derived from glucose through the malate aspartate shuttle is the specific signal underlying insulin secretion after being stimulated via cAMP [12]. It has been shown that the cAMP/PKA pathway potentiates the release of insulin via increased effectiveness of KATP-channel-independent actions of glucose [29], which is consistent with our observations. Intracellular calcium is unaffected by PKA activation, and PKA effects on insulin secretion are mediated by the phosphorylation of various downstream proteins [30]. The mechanisms leading to the rise in insulin secretion are studied with electrophysiological and optical methods which monitor the movements and exocytosis of individual insulin granules [31,32]. Changes in the size of distinct granular pools, facilitation of granule recruitment from the pools to the plasma membrane, and the acceleration of the priming process that confers granules with release competence may all play a role as well as the involvement of SNARE complexes. Relating this back to our study, the PKA pathway stemming from the increase in cAMP ties to the amplification pathway via glucose metabolism and is most likely a key player in insulin release at such high glucose concentrations.

Protein kinase A (PKA) has a permissive role in increasing insulin secretion that has

In addition to what we have shown, there are other glucose-associated shunts that lead to insulin secretion through the amplification pathway. Our data suggest that glycolysis

itself maintains glucose dependence well above 24 mM glucose. Although, other than direct intermediates of glucose metabolism, there are potential excess-fuel detoxification pathways dealing with glycerol and free fatty acid formation and their extracellular release [32]. Free fatty acids play a role in insulin secretion by stimulating monoacylglycerol formation, whereas the inhibition of monoacylglycerol lipase activity decreases insulin secretion [33]. The aforementioned pathways may include the diversion of glucose carbons to triglycerides and cholesterol esters. Aspects relating to mitochondrial energy metabolism independent signals, including 1-monoglycerol, diacylglycerol, and malonyl-CoA are pieces of the amplification pathway to explore at high glucose levels. [33]. Future studies should investigate glycerol release and free fatty acids in extreme glucose conditions to further understand these intricate pathways.

#### *3.3. Paracrine Effects of Extreme Glucose via Alpha Cells*

Increasing glucose levels have been shown to lead to the induction of cAMP oscillations in both alpha and beta cells [34]. The alpha cells in pancreatic islets secrete GLP-1, which generally suppresses glucagon secretion [35]. Glucose-related glucagon secretion is observed in islets and reflects direct effects on alpha cells [29,30]. L-arginine can potently stimulate GLP-1 release in islets, and there is evidence that glucose may potentiate L-arginine-stimulated insulin secretion via PKA [36]. The relationship between glucose and GLP-1R is pivotal to understanding how glucose leads to an increase in cAMP.

It is known that glucose inhibits glucagon secretion by lowering cytoplasmic calcium in the alpha-cell; however, stimulation of glucagon at high glucose concentrations does not require an increase in intracellular calcium, and at higher glucose, glucagon secretion is actually stimulated. This paradoxical stimulation of glucagon release occurs around at least 25–30 mM [30,31]. In fact, high glucose has been shown to have a stimulatory effect on glucagon secretion possibly exceeding that of the inhibitory influence [37–39]. Looking at Figure 6, our studies demonstrate the connection of increasing glucose concentrations to increased GLP-1 and increased glucagon secreted by alpha cells. Our study suggests that the increase in incretins and glucagon from the alpha cells at high glucose concentrations could act on the beta cell to augment insulin secretion through the GLP-1 receptor pathway at these extreme levels.

#### *3.4. Clinical Relevance*

Diabetic patients have survived extreme glucose levels of over 100 mM [2,4,10]. Our study shows that insulin secretion occurs at extreme glucose levels, but as the concentration of glucose in the blood increases, insulin secretion in the higher glucose range does not keep to the same rate as in lower glucose concentrations, which is evidenced by the decrease in slope (Figure 1). However, this ability of beta cells to secrete insulin in these extremes of hyperglycemia is what distinguishes hyperosmolar hyperglycemic nonketoic syndrome from diabetic ketoacidosis. There is enough insulin present to prevent ketosis but not sufficient insulin to stimulate glucose utilization in target tissues (~10× as much insulin needed) [40]. We also observed that the extreme osmolarity increase associated with extreme hyperglycemia does not appear to negatively impact insulin secretion, at least in our in vitro studies in mouse islets. Thus, although individuals have survived, their bodies endured extreme stress during these instances. It should be noted that glucotoxicity has an effect on limiting the body's ability to secrete insulin in extreme conditions, but this is considered a more chronic state than what we report.

Overall, we showed that insulin secretion from islets of both mice and human donors continues to increase in a dose-dependent manner to much higher glucose levels than previously thought. It is possible that novel pathways to insulin secretion could be identified only by stimulation in extremely high glucose. Once identified, it may be possible to develop novel therapeutics that could stimulate this secretory activity without requiring extremely high glucose.

#### *3.5. Strengths and Limitations*

An important strength of this study was the consistent observation in both murine and human islets in multiple trials which showed that the dose-dependent range of glucosestimulated insulin secretion extends far higher than commonly thought. We further show that these increases rely on increases in glycolytic activity and cAMP, but not on changes in intracellular calcium. In addition, these studies show that osmolarity does not impact insulin secretion in vitro, which eliminates a potential confounding variable. Limitations in this study include the fact that isolated islets in vitro lack normal neural and humoral inputs found in vivo that can modulate function. Islets also lack the vasculature of their in vivo environment, which can impact how nutrients like glucose reach the islet. These are issues common to any in vitro study of pancreatic islets. Lastly, although our study shows an important role for cAMP in the insulin response to extreme glucose, many other factors of the amplification pathway could also be involved. Examining additional mechanisms will be the focus of future work.

#### **4. Materials and Methods**

#### *4.1. Islet Sources and Isolation*

Mouse islets were isolated from male CD-1 (Envigo, Indianapolis, IN, USA) mice ages 8–12 weeks, as previously described [41]. Briefly, pancreatic islets were isolated using collagenase-*p* digestion (Roche Diagnostics, Indianapolis, IN, USA) followed by centrifugation using Histopaque 1100 (Sigma-Aldrich, St. Louis, MO, USA). Islets were allowed to recover overnight in RPMI 1640 (Invitrogen, Carlsbad, CA, USA), supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin before being used for experiments. All animal procedures were approved by the Ohio University Institutional Animal Care and Use Committee. Human islets from deidentified donors were obtained from the University of Alberta IsletCore and the University of Alberta/Alberta Health Services Clinical Islet Laboratory.

#### *4.2. Calcium Imaging*

Fura-2 AM fluorescence imaging was utilized to measure intracellular calcium levels. Perifused solutions first passed through an inline heater to a temperature of 35+/−3 degrees Celsius into an open diamond bath imaging chamber (Warner Instruments, Cat: 64-0288) which was mounted using a stage adapter (Warner Instruments, Cat: 64-0298). Observation of islets was performed using a Hamamatsu ORCA-Flash4.0 digital camera (Hamamatsu Photonics K.K., Hamamatsu City, Japan, Model C11440-22CU) mounted on a BX51WIF fluorescence microscope with a 10X objective (Olympus, Tokyo, Japan). Excitation light was provided by a xenon burner supplied to the image field through a light pipe and filter wheel (Sutter Instrument Co., Novato, CA, USA, Model LB-LS/30) with a Lambda 10-3 Optical Controller (Sutter Instrument Co., Novato, CA, USA, Model LB10- 3-1572). Images were taken sequentially with 340 nm and 380 nm excitation to produce each ratio from emitted light at 510 nm. Data were analyzed using cellSens Dimension 1.13 imaging software (Olympus, Tokyo, Japan) [42].

Islets were exposed to KRB containing 1 µM fura-2 AM and incubated for 30 min. They then were transferred onto the calcium scope. The fura-2 signal was recorded for two different experimental protocols. Protocol 1: 3 mM glucose (G) for 5 min, 24 G for 15 min, and 84 G for 15 min. Protocol 2: 3 mM glucose (G) for 5 min, 24 G for 15 min, 24 G containing 250 µM tolbutamide for 15 min, and back to 3 G for 25 min.

#### *4.3. Insulin Secretion*

To study insulin secretion, glucose-stimulated insulin secretion (GSIS) assays were performed. All GSIS used 12-well plates with 20 islets per well in 1 mL of KRB solution. Islets were size matched to aid in normalization as discussed in [43]. Briefly, islets were placed in 0 mM glucose for one hour, then transferred to the experimental conditions for an additional hour. Supernatants were collected from experimental conditions of 0,

12, 24, 36, 48, 60, 72, 84, and 144 mM. Using the same approach, we studied additional conditions using osmolarity-matched solutions at ~425 mOsm. Osmolarity was measured with Wescor Vapor Pressure Osmometer (Model 5520). Specific stimulators or inhibitors were placed into 24 G and 84 G to examine effects of exendin (10 nM, Sigma Aldrich, St. Louis, MO, USA, [18]), forskolin (10 µM, Sigma Aldrich, St. Louis, MO, USA, [44]), MDL-12330A (10 µM, Sigma Aldrich, St. Louis, MO, USA, [45]), and Ro-28-1675 (500 nM, Axon Medchem (Reston, VA, USA)). Insulin secretion was measured using mouse (Cat#80- INSMSU-E10) and human (Cat#80-INSHU-E01.1) ELISA following the manufacturer's directions (ALPCO, Salem, NH, USA). Intra-assay variability was kept to below 15% for all studies. D-mannitol (Sigma-Aldrich, St. Louis, MO, USA) was used to balance the osmolarity of Modified KRB solutions to ~425 mOsm.

#### *4.4. Cell Death Quantification*

Islets were incubated in a glucose solution for 48 h in standard RPMI media supplemented 10% fetal bovine serum and 1% penicillin/streptomycin. Twenty islets were placed in each well per treatment in a 12-well plate. The 12-well plate contained the following treatments. Standard RPMI media, RPMI media+144 mM mannitol, RPMI media+60 mM glucose+84 mM mannitol, and finally RPMI+144 mM glucose. Cell death was measured with propidium iodide (Sigma-Aldrich, St. Louis, MO, USA) and annexin V (Invitrogen, Carlsbad, CA, USA) staining. Apoptosis was measured using annexin V (488 nm excitation/525 nm emission), which detects phosphatidylserine when it is exposed to the outer leaflet of the plasma membrane during apoptosis [46]. Propidium iodide (535 nm excitation/620 nm emission), which is a cell exclusion dye, was used to detect generalized cell death. Regions of interest were drawn around islets to measure fluorescence intensity per islet for each individual islet normalized to surface area. These techniques have been used in previous publications [47,48] including in comparison to other methods to measure cell death [49].

#### *4.5. Statistical Analysis*

Statistical analysis was performed using R Statistical Computing Software. Data are expressed as the mean ± standard error of the mean. Data were tested for normality using Shapiro-Wilk test and for equal variance using Levene's test. Henze-Zirkler's multivariate normality test was used for the insulin correlation data. All comparisons were analyzed using two-tailed *t*-test for comparisons of two groups or one-way ANOVA with Tukey's post hoc test for more than two groups. Differences between groups were considered significant at *p* < 0.05. Spearman's rank correlation coefficient was used to analyze insulin secretion patterns in Figure 1.

#### *4.6. Ethical Approval*

All animal procedures were approved by the Ohio University Institutional Animal Care and Use Committee. Human islet isolation was approved by the Human Research Ethics Board at the University of Alberta (Pro00013094). All donors' families gave informed consent for the use of pancreatic tissue in research. Donor information was deidentified prior to our acquisition.

**Author Contributions:** Conceptualization, C.S.N., D.R.R., K.L.C., K.M.G.; methodology, all authors; software, N.B.W., C.S.N.; validation, all authors; formal analysis, K.M.G., C.S.N., N.B.W.; investigation, K.M.G., C.S.N., N.B.W., W.J.K.; resources, K.M.G., C.S.N.; data curation, K.M.G., C.S.N., N.B.W.; writing—original draft preparation, K.M.G., N.B.W.; writing—review and editing, C.S.N., K.M.G., N.B.W.; visualization, K.M.G., N.B.W., C.S.N.; supervision, C.S.N.; project administration, C.S.N.; funding acquisition, C.S.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by NIDDK R15 DK121247, the Diabetes Institute, and the Ohio University Heritage College of Osteopathic Medicine.

**Institutional Review Board Statement:** All animal procedures were approved by the Ohio University Institutional Animal Care and Use Committee. Human islet isolation was approved by the Human Research Ethics Board at the University of Alberta (Pro00013094 and Pro00001620). All donors' families gave informed consent for the use of pancreatic tissue in research. Donor information was deidentified prior to our acquisition.

**Informed Consent Statement:** All donors' families gave informed consent for the use of pancreatic tissue in research.

**Data Availability Statement:** All data used to support the findings of this study are available from the corresponding author upon request.

**Acknowledgments:** Human islets were provided by the Alberta Islet Distribution Program and by the Alberta Diabetes Institute IsletCore, University of Alberta in Edmonton. Islet isolation was approved by the Human Research Ethics Board at the University of Alberta (Pro00013094). The graphical abstract and Figure 6 were made using BioRender.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Uma D. Kabra 1,2 , Charles Affourtit <sup>3</sup> and Martin Jastroch 4,5,\***


**Abstract:** The development of obesity and type 2 diabetes (T2D) has been associated with impaired mitochondrial function. In pancreatic beta (β) cells, mitochondrial energy metabolism plays a central role in triggering and controlling glucose-stimulated insulin secretion (GSIS). Here, we have explored whether mitochondrial bioenergetic parameters assessed with Seahorse extracellular flux technology can quantitatively predict insulin secretion. We metabolically stressed male C57BL/6 mice by high-fat feeding (HFD) and measured the glucose sensitivity of islet respiration and insulin secretion. The diet-induced obese (DIO) mice developed hyperinsulinemia, but no pathological secretory differences were apparent between isolated DIO and chow islets. Real-time extracellular flux analysis, however, revealed a lower respiratory sensitivity to glucose in DIO islets. Correlation of insulin secretion with respiratory parameters uncovers compromised insulin secretion in DIO islets by oxidative power. Normalization to increased insulin contents during DIO improves the quantitative relation between GSIS and respiration, allowing to classify dysfunctional properties of pancreatic insulin secretion, and thereby serving as valuable biomarker for pancreatic islet glucose responsiveness and health.

**Keywords:** mitochondria; bioenergetics; glucose-stimulated insulin secretion; pancreatic islets; respiration

#### **1. Introduction**

Pancreatic β cells are specialized endocrine cells that integrate signals from glucose and other fuels to control the secretion of insulin [1]. Glucose induces insulin secretion via both triggering and amplifying pathways [2]. The triggering pathway involves oxidative glucose catabolism, a rise in the cytosolic adenosine triphosphate/adenosine diphosphate (ATP/ADP) ratio, closure of ATP-sensitive potassium (KATP) channels, depolarization of the plasma membrane, influx of calcium ions, and eventual exocytosis of insulin-containing granules [3–5]. The amplifying pathways boost this glucose-stimulated insulin secretion (GSIS) in a KATP-channel-independent (but Ca2+-dependent) way [4,6]. Many steps of the β cell insulin secretory pathways can affect the efficiency of insulin secretion. Mitochondria are intimately involved in glucose catabolism because oxidative phosphorylation has control over the ATP/ADP ratio and, thus, GSIS [7]. Oxidative phosphorylation is ideally assessed by combining respiratory flux and mitochondrial membrane potential measurements [8,9]. While mitochondrial membrane potential measurements of β cell mitochondria are challenging in intact β cells because of glucose-evoked plasma membrane

**Citation:** Kabra, U.D.; Affourtit, C.; Jastroch, M. Respiratory Parameters for the Classification of Dysfunctional Insulin Secretion by Pancreatic Islets. *Metabolites* **2021**, *11*, 405. https:// doi.org/10.3390/metabo11060405

Academic Editors: Melkam Kebede and Belinda Yau

Received: 3 June 2021 Accepted: 18 June 2021 Published: 21 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

potential fluctuations [10], cellular respiratory flux measurements are relatively straightforward [11]. Several parameters are readily calculated from whole-cell oxygen uptake to give insight in the efficiency by which energy liberated from oxidative glucose breakdown is conserved as ATP. These parameters include glucose sensitivity (GS), i.e., the magnitude of respirational increase upon glucose stimulus, and coupling efficiency (CE) of oxidative phosphorylation, i.e., the part of mitochondrial respiration coupled to ATP synthesis [8]. Notably, these bioenergetic parameters are internally normalized and, thus, dimensionless, which renders them more sensitive indicators of oxidative phosphorylation than absolute oxygen uptake rates [8]. For example, CE has been instrumental for demonstrating a glucose sensitivity of oxidative phosphorylation in INS-1E insulinoma cells that is regulated by mitochondrial uncoupling protein-2 [12]. This bioenergetic regulation is reflected by insulin secretion activity, which suggests the possibility that CE has broader predictive power that may be exploited to forecast β cell dysfunction. forward [11]. Several parameters are readily calculated from whole-cell oxygen uptake to give insight in the efficiency by which energy liberated from oxidative glucose breakdown is conserved as ATP. These parameters include glucose sensitivity (GS), i.e., the magnitude of respirational increase upon glucose stimulus, and coupling efficiency (CE) of oxidative phosphorylation, i.e., the part of mitochondrial respiration coupled to ATP synthesis [8]. Notably, these bioenergetic parameters are internally normalized and, thus, dimensionless, which renders them more sensitive indicators of oxidative phosphorylation than absolute oxygen uptake rates [8]. For example, CE has been instrumental for demonstrating a glucose sensitivity of oxidative phosphorylation in INS-1E insulinoma cells that is regulated by mitochondrial uncoupling protein-2 [12]. This bioenergetic regulation is reflected by insulin secretion activity, which suggests the possibility that CE has broader predictive power that may be exploited to forecast β cell dysfunction. In this study, we correlated mitochondrial respiration and insulin secretion in pan-

measurements [8,9]. While mitochondrial membrane potential measurements of β cell mitochondria are challenging in intact β cells because of glucose-evoked plasma membrane potential fluctuations [10], cellular respiratory flux measurements are relatively straight-

*Metabolites* **2021**, *11*, x FOR PEER REVIEW 2 of 7

In this study, we correlated mitochondrial respiration and insulin secretion in pancreatic islets from chow-fed and diet-induced obese (DIO) mice, with the aim to identify parameters that allow classification of insulin secretion deficiencies. DIO mice represent a model of a well-established risk scenario in which obesity leads to insulin resistance, β cell dysfunction, and eventually type 2 diabetes mellitus [13]. Our findings suggest that loss of mitochondrial respiratory sensitivity to glucose is an early warning sign for compromised insulin secretion. creatic islets from chow-fed and diet-induced obese (DIO) mice, with the aim to identify parameters that allow classification of insulin secretion deficiencies. DIO mice represent a model of a well-established risk scenario in which obesity leads to insulin resistance, β cell dysfunction, and eventually type 2 diabetes mellitus [13]. Our findings suggest that loss of mitochondrial respiratory sensitivity to glucose is an early warning sign for compromised insulin secretion. **2. Results and Discussion** 

#### **2. Results and Discussion** *2.1. Inducing Metabolic Stress in Mice*

#### *2.1. Inducing Metabolic Stress in Mice* We tested our hypothesis on the relation between respiration and insulin secretion

We tested our hypothesis on the relation between respiration and insulin secretion in DIO mice, a typical model for the development of insulin resistance and hyperinsulinemia. Male C57BL/6 mice were kept on high-fat diet (HFD) for 16 weeks to induce obesity (~45.8 g HFD vs. ~28.8 g chow; Figure 1A). Plasma insulin levels were dramatically increased (Figure 1B). However, plasma glucose levels were unchanged (Figure 1C), suggesting that hyperinsulinemia was sufficient to compensate insulin resistance in HFD mice. in DIO mice, a typical model for the development of insulin resistance and hyperinsulinemia. Male C57BL/6 mice were kept on high-fat diet (HFD) for 16 weeks to induce obesity (~45.8 g HFD vs. ~28.8 g chow; Figure 1A). Plasma insulin levels were dramatically increased (Figure 1B). However, plasma glucose levels were unchanged (Figure 1C), suggesting that hyperinsulinemia was sufficient to compensate insulin resistance in HFD mice.

**Figure 1.** Characterization of chow (black bars) and HFD (white bars) fed C57BL/6 mice. (**A**) Body weight; (**B**) plasma insulin level; and (**C**) plasma glucose levels. The mice were 8 weeks old at the start of the feeding experiment. Data are represented as ± SEM (*n* = 8 mice per group). Statistical significance of mean differences was tested by unpaired two-tailed student *t*-test. *p* < 0.001 (\*\*\*). **Figure 1.** Characterization of chow (black bars) and HFD (white bars) fed C57BL/6 mice. (**A**) Body weight; (**B**) plasma insulin level; and (**C**) plasma glucose levels. The mice were 8 weeks old at the start of the feeding experiment. Data are represented as ± SEM (*n* = 8 mice per group). Statistical significance of mean differences was tested by unpaired two-tailed student *t*-test. *p* < 0.001 (\*\*\*).

#### *2.2. Respiratory Activity but Not Insulin Secretion Is Impaired in Islets of DIO Mice 2.2. Respiratory Activity but Not Insulin Secretion Is Impaired in Islets of DIO Mice*

Interestingly, the assessment of insulin secretion between chow and DIO islets revealed no differences (Figure 2A). In contrast, averaged real-time oxygen consumption traces in Figure 2B demonstrate clearly that respiration in DIO islets differs drastically Interestingly, the assessment of insulin secretion between chow and DIO islets revealed no differences (Figure 2A). In contrast, averaged real-time oxygen consumption traces in Figure 2B demonstrate clearly that respiration in DIO islets differs drastically from that exhibited by their control counterparts. After normalizing oxygen uptake to DNA content and correcting respiratory activity for non-mitochondrial (i.e., rotenone-andantimycin-A-resistant) oxygen consumption, it transpired that there was no difference in basal respiration between the two systems (Figure 2C). However, the strong respiratory

stimulation provoked by glucose in control islets was more than halved in DIO islets (Figure 2D,E), which was the combined result of attenuated respiration linked to mitochondrial proton leak (Figure 2F) and ATP synthesis (Figure 2G). In other words, both oligomycin-sensitive and -insensitive oxygen uptake was decreased by DIO, which explains why the CE of oxidative phosphorylation, which reflects the oligomycin sensitivity of overall glucose-stimulated respiration (GSR), was not affected to a statistically significant degree (Figure 2H). The reduction of both proton leak and ATP-linked respiration can be indicative of compromised substrate oxidation capacity [8]. Next, we plotted absolute secreted insulin values vs respiratory parameters related to insulin triggering (ATP-linked respiration and GSR), and assessed their relation by correlation analysis. Within chow islets, insulin values correlate significantly with ATP-linked respiration and GSR (Figure 2I,J). This relationship appeared to be shifted upwards in DIO islets (dotted regression line). In our experimental conditions in vitro, however, no secretagogues or other amplifying mechanisms are present, to the best of our knowledge, which could selectively act on DIO islets. However, we cannot formally exclude amplifying factors deriving from glucose catabolism [14], and we thus refer to altered secretory pathways of insulin secretion rather than to altered triggering of insulin secretion. and-antimycin-A-resistant) oxygen consumption, it transpired that there was no difference in basal respiration between the two systems (Figure 2C). However, the strong respiratory stimulation provoked by glucose in control islets was more than halved in DIO islets (Figure 2D,E), which was the combined result of attenuated respiration linked to mitochondrial proton leak (Figure 2F) and ATP synthesis (Figure 2G). In other words, both oligomycin-sensitive and -insensitive oxygen uptake was decreased by DIO, which explains why the CE of oxidative phosphorylation, which reflects the oligomycin sensitivity of overall glucose-stimulated respiration (GSR), was not affected to a statistically significant degree (Figure 2H). The reduction of both proton leak and ATP-linked respiration can be indicative of compromised substrate oxidation capacity [8]. Next, we plotted absolute secreted insulin values vs respiratory parameters related to insulin triggering (ATP-linked respiration and GSR), and assessed their relation by correlation analysis. Within chow islets, insulin values correlate significantly with ATP-linked respiration and GSR (Figure 2I,J). This relationship appeared to be shifted upwards in DIO islets (dotted regression line). In our experimental conditions in vitro, however, no secretagogues or other amplifying mechanisms are present, to the best of our knowledge, which could selectively act on DIO islets. However, we cannot formally exclude amplifying factors deriving from glucose catabolism [14], and we thus refer to altered secretory pathways of insulin secretion rather than to altered triggering of insulin secretion.

from that exhibited by their control counterparts. After normalizing oxygen uptake to DNA content and correcting respiratory activity for non-mitochondrial (i.e., rotenone-

*Metabolites* **2021**, *11*, x FOR PEER REVIEW 3 of 7

**Figure 2.** Glucose-stimulated insulin secretion and mitochondrial bioenergetics of pancreatic islets from chow and HFDfed mice. (**A**) Insulin secretion. Batches of eight size-matched islets were exposed to either 2 or 16.5 mM glucose for 1 h before supernatants and lysates were collected. (**B**) Representative time-resolved oxygen consumption traces. Batches of 30 size-matched islets were exposed to 2 mM glucose to assess basal respiration (4 cycles), glucose-stimulated respiration (10 cycles), proton leak respiration by inhibition of ATP synthase using oligomycin (10 cycles), and non-mitochondrial respiration by final injection of rotenone/antimycin A. (**C**) Mitochondrial basal respiration at low glucose (2 mM). (**D**) Glucose-stimulated respiration expressed as percentage of basal (**E**) Mitochondrial stimulated respiration at high glucose (16.5 mM). (**F**) Proton leak respiration at high glucose. (**G**) ATP-linked respiration at high glucose. (**H**) Coupling efficiency at high glucose. (**I**) Correlation of insulin secretion (absolute values) and ATP-linked respiration. (**J**) Correlation of insulin secretion (absolute values) and glucose-stimulated respiration. Data are represented as means ± SEM (*n* = 6 mice **Figure 2.** Glucose-stimulated insulin secretion and mitochondrial bioenergetics of pancreatic islets from chow and HFD-fed mice. (**A**) Insulin secretion. Batches of eight size-matched islets were exposed to either 2 or 16.5 mM glucose for 1 h before supernatants and lysates were collected. (**B**) Representative time-resolved oxygen consumption traces. Batches of 30 size-matched islets were exposed to 2 mM glucose to assess basal respiration (4 cycles), glucose-stimulated respiration (10 cycles), proton leak respiration by inhibition of ATP synthase using oligomycin (10 cycles), and non-mitochondrial respiration by final injection of rotenone/antimycin A. (**C**) Mitochondrial basal respiration at low glucose (2 mM). (**D**) Glucose-stimulated respiration expressed as percentage of basal (**E**) Mitochondrial stimulated respiration at high glucose (16.5 mM). (**F**) Proton leak respiration at high glucose. (**G**) ATP-linked respiration at high glucose. (**H**) Coupling efficiency at high glucose. (**I**) Correlation of insulin secretion (absolute values) and ATP-linked respiration. (**J**) Correlation of insulin secretion (absolute values) and glucose-stimulated respiration. Data are represented as means ± SEM (*n* = 6 mice per group, each mouse was considered as independent experiment and islets were plated in triplicate). Statistical significance of mean differences was tested by unpaired two-tailed student *t*-test. *p* < 0.05 (\*), *p* < 0.001 (\*\*\*).

#### *2.3. Compensating Increase of Insulin Content in DIO Islets Masks Defects of the Triggering Pathway 2.4. Classifying Defects in Insulin Secretion*  From Figures 3C–E, it transpires that secreted insulin values require normalization

*Metabolites* **2021**, *11*, x FOR PEER REVIEW 4 of 7

per group, each mouse was considered as independent experiment and islets were plated in triplicate). Statistical signifi-

cance of mean differences was tested by unpaired two-tailed student *t*-test. *p* < 0.05 (\*),*p* < 0.001 (\*\*\*).

to fix the secretory impairment.

*Pathway* 

Exploring compensatory responses to loss of insulin sensitivity in HFD mice (Figure 1), we found a 15% higher insulin content in DIO islets (Figure 3A), consistent with increases in β cell proliferation during DIO [15]. Normalizing insulin secretion to insulin content, GSIS of the DIO islets was lower than that from control islets (Figure 3B). Importantly, plotting GSIS values normalized to insulin content shifted DIO insulin values onto the regression of chow islets, with a better regression to glucose-stimulated than to ATP-linked respiration (Figure 3C,D). Insulin secretion correlates less well with coupling efficiency (CE, Figure 3E), but it should be noted that the internal standardization benefits data comparisons between different experimental settings and laboratories. The normalizationdependency of the GSIS phenotype suggests that DIO islets respond to an obesity-induced drop in glucose sensitivity by increasing insulin content, a response that appears to fix the secretory impairment. to insulin content to establish a robust linear relationship between insulin and bioenergetics parameters. The relationship between oxidative phosphorylation and GSIS leads to a simple linear correlation model to classify defects of insulin secretion (Figure 3F). In relation to control values, descending along the regression line suggests reduced oxidative power by either compromised substrate delivery or respiratory dysfunction. In the case of DIO islets, the secretory pathway is compromised by reduced oxidative power, which could be mediated by impaired glucose uptake, glycolysis, or pyruvate oxidation—the latter being recently suggested in response to impaired mitochondrial dynamics [16] or inflammation [17]. In contrast, ascending values suggest improved substrate delivery or oxidative phosphorylation. Upwards deviation from regression is explained by amplifying pathways, while downwards deviation points towards secretory dysfunction downstream or no mitochondrial impact.

*2.3. Compensating Increase of Insulin Content in DIO Islets Masks Defects of the Triggering* 

Exploring compensatory responses to loss of insulin sensitivity in HFD mice (Figure 1), we found a 15% higher insulin content in DIO islets (Figure 3A), consistent with increases in β cell proliferation during DIO [15]. Normalizing insulin secretion to insulin content, GSIS of the DIO islets was lower than that from control islets (Figure 3B). Importantly, plotting GSIS values normalized to insulin content shifted DIO insulin values onto the regression of chow islets, with a better regression to glucose-stimulated than to ATPlinked respiration (Figure 3C,D). Insulin secretion correlates less well with coupling efficiency (CE, Figure 3E), but it should be noted that the internal standardization benefits data comparisons between different experimental settings and laboratories. The normalization-dependency of the GSIS phenotype suggests that DIO islets respond to an obesityinduced drop in glucose sensitivity by increasing insulin content, a response that appears

**Figure 3.** Glucose-stimulated insulin secretion in chow and HFD mice. Batches of eight size-matched islets were exposed to either 2 or 16.5 mM glucose for 1 h at which point supernatants and lysates were collected for (**A**) insulin content, and (**B**) secreted insulin measurements. Correlation between insulin secretion (% content) and (**C**) glucose-stimulated respiration, (**D**) ATP-linked respiration, and (**E**) coupling efficiency (**F**) correlation model classifying defects of insulin secretion. Data are represented as means ± SEM (*n* = 6 mice per group, each mouse was considered as independent experiment and islets were plated in triplicate). Statistical significance of mean differences was tested by unpaired two-tailed student *t*-test. *p* < 0.05 (\*), *p* < 0.001 (\*\*\*).

#### *2.4. Classifying Defects in Insulin Secretion*

From Figure 3C–E, it transpires that secreted insulin values require normalization to insulin content to establish a robust linear relationship between insulin and bioenergetics parameters. The relationship between oxidative phosphorylation and GSIS leads to a simple linear correlation model to classify defects of insulin secretion (Figure 3F). In relation to control values, descending along the regression line suggests reduced oxidative power by either compromised substrate delivery or respiratory dysfunction. In the case of DIO islets, the secretory pathway is compromised by reduced oxidative power, which could be mediated by impaired glucose uptake, glycolysis, or pyruvate oxidation—the

latter being recently suggested in response to impaired mitochondrial dynamics [16] or inflammation [17]. In contrast, ascending values suggest improved substrate delivery or oxidative phosphorylation. Upwards deviation from regression is explained by amplifying pathways, while downwards deviation points towards secretory dysfunction downstream or no mitochondrial impact.

#### **3. Materials and Methods**

Animals—male C57BL/6 mice with an age of 8 to 10 weeks were purchased from Janvier Labs (Le Genest-Saint-Isle, France). The animals were maintained on a 12/12 h light/dark cycle in a temperature-controlled environment and allowed free access to standard chow diet (5.6% fat, LM-485, Harlan Teklad) or a high-fat diet (HFD) (58% kcal fat; Research Diets Inc., New Brunswick, NJ, USA) for 16 weeks. All in vivo procedures were conducted under the guidelines of the Institutional Animal Care Committee of the Helmholtz Center Munich, which approved all animal maintenance and experimental procedures. The animal experiments complied with all ethical regulations for animal testing and research, including animal maintenance and experimental procedures that the animal welfare authorities of the local animal ethics committee of the state of Bavaria (Regierung Oberbayern) approved in accordance with European guidelines.

Islet isolation and culture—mouse islets were isolated by digestion with collagenase as described elsewhere [12]. Around 150–200 islets were obtained per mouse. Islets were incubated overnight in RPMI 1640 culture medium supplemented with 10% (*v/v*) fetal calf serum (Life technologies) at 37 ◦C and 5% CO<sup>2</sup> before experimentation.

Insulin secretion—groups of eight size-matched islets were handpicked into individual wells of V-bottomed 96-well plates and incubated for 60 min at 37 ◦C in HEPES-balanced Krebs-Ringer (KRH) bicarbonate buffer containing 114 mM NaCl, 4.7 mM KCl, 2.5 mM CaCl2, 1.16 mM MgSO4, 1.2 mM KH2PO4, 25.5 mM NaHCO3, 20 mM HEPES (pH 7.2–7.4), supplemented with 0.2% (*w/v*) bovine serum albumin (BSA), and 2 mM glucose. The islets were then incubated for a further 60 min in KRH-bicarbonate buffer containing 2 or 16.5 mM glucose. Subsequently, supernatants were collected to quantify secreted insulin. Following the secretion assay, islets were lysed with ice-cold RIPA buffer to allow total insulin content to be determined. Insulin was detected with the Ultra-Sensitive Mouse Insulin Elisa Kit (ALPCO, Salem, NH, USA). Data were normalized to DNA content measured with the Quant-it Pico Green DNA assay kit (Invitrogen, Darmstadt, Germany).

Mitochondrial respiration—oxygen consumption rates (OCR) were measured in isletcapture plates of the XF24 extracellular flux analyzer (Agilent, Seahorse Bioscience, Santa Clara, CA, USA). Briefly, groups of 30 size-matched islets were handpicked into individual wells of islet capture plates and incubated for 60 min at 37 ◦C without CO<sup>2</sup> in KRHbicarbonate free buffer containing 2 mM glucose. Additional glucose (16.5 mM), oligomycin (10 µg/mL), and a mixture of rotenone and antimycin A (both 2 µM) were injected sequentially. Mitochondrial respiration was calculated by subtracting non-mitochondrial respiration from all other oxygen uptake rates. The individual bioenergetics parameters of OXPHOS parameters were calculated as follows: basal mitochondrial respiration = (last rate measured before glucose and/or other secretagogues injection) − (non-mitochondrial respiration rate). Glucose-stimulated mitochondrial respiration = (last rate measured before oligomycin injection) − (non-mitochondrial respiration rate). Proton-leak-linked respiration = (minimum rate measured after oligomycin injection) − (non-mitochondrial respiration rate). ATP-synthesis-linked respiration = (glucose-stimulated mitochondrial respiration) − (proton leak-linked respiration). Coupling efficiency = the fraction of respiration used to drive ATP synthesis for each run, calculated as CE = 1 − (proton-leaklinked respiration/mitochondrial glucose-stimulated respiration). Data were normalized to DNA content.

Metabolic tests—mice were fasted for six hours prior to blood collection. Blood samples taken from the tail vein were used to measure glycaemia with a glucometer (Abbott, Wiesbaden, Germany) and insulin by ELISA.

Statistical analysis—statistical analysis was performed with GraphPad Prism version 6.0. The data passed normality tests (Shapiro–Wilk and Kolmogorov–Smirnov test) and the groups were compared using unpaired Student's *t* test. All the data are shown as mean ± standard error of mean (S.E.M). *p* values < 0.05 were considered statistically significant.

## **4. Conclusions**

In this study, we demonstrate that mitochondrial respiratory parameters have predictive value for insulin secretion from pancreatic β cells. In a typical model for metabolic disease, DIO mice, we found that defective insulin secretion is compensated by increased insulin content. Although absolute values of insulin secretion are not affected, mitochondrial respiration is severely compromised in DIO islets. Correlation of respiration and GSIS is firmly established by normalizing GSIS to insulin content, showing that mitochondrial respiratory parameters quantitatively predict changes in GSIS. Absolute values of ATP-linked respiration and GSR predict insulin release the best, while CE is less predictive. CE, however, is a powerful bioenergetic parameter that has been successfully applied to uncover molecular mechanisms in mitochondria, e.g., the role of UCP2 in β cells [8]. CE is based on the thermodynamic laws of energy conversion that mitochondrial oxidation energy is either converted to ATP or lost as heat due to the mitochondrial proton leak. CE is defined as the fraction of energy that is converted to ATP (thus ranging from 0 to 1). As internally standardized parameter, CE is not prone to variation in absolute values between independent experiments. Although CE correlates with GSIS, the relation is steeper and more variable concerning linear regression, as compared to ATP-linked respiration and GSR in our experimental setup. Thus, CE may only be used to compare independent studies. However, the relation of CE and GSIS suggests a "threshold" CE value that is required to trigger insulin secretion. Considering previous studies from our laboratory [16,18], we find a tight CE of about 0.4–0.6 as requirement for insulin triggering in islets and in β cell models. Based on the results we designed a model correlating GSR or ATP linked respiration vs. insulin secretion that allows classifying dysfunctional properties of pancreatic insulin secretion under pathological conditions. All these analyses suggest that mitochondrial bioenergetic parameters reflect insulin secretion in a quantifiable manner, and may thus serve as biomarkers for glucose responsiveness and pancreatic islet health.

**Author Contributions:** Conceptualization, M.J., C.A., and U.D.K.; investigation, U.D.K.; resources, M.J.; writing—original draft preparation, U.D.K., C.A., and M.J.; writing—review and editing, U.D.K., M.J., and C.A; visualization, U.D.K.; supervision, M.J.; project administration, M.J.; funding acquisition, M.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was partially supported by the German Center for Diabetes Research (DZD) and Novo Nordisk Fonden (to M. J. grant number 0059646).

**Institutional Review Board Statement:** All in vivo procedures were conducted under the guidelines of the Institutional Animal Care Committee of the Helmholtz Center Munich, which approved all animal maintenance and experimental procedures. The ethic code 5.1-568-Gas was granted by the Landratsamt München. The animal experiments complied with all ethical regulations for animal testing and research, including animal maintenance and experimental procedures, which the animal welfare authorities of the local animal ethics committee of the state of Bavaria (Regierung Oberbayern) approved, in accordance with European guidelines.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors thank Katrin Pfuhlmann and Maria Kutschke at the Helmholtz Zentrum München for excellent technical assistance. We are grateful for excellent support by the central animal facility of the Helmholtz Center Munich and their animal caretakers.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

## **References**

