Thiazides Attenuate Insulin Secretion Through Inhibition of Mitochondrial Carbonic Anhydrase 5b in β-Islet Cells in Mice

Abstract

Significance Statement

Thiazide diuretics (thiazides) are among the most widely prescribed drugs worldwide, but their use is associated with glucose intolerance and new-onset diabetes mellitus. The molecular mechanisms remain elusive. Our study reveals that thiazides attenuate insulin secretion through inhibition of the mitochondrial carbonic anhydrase isoform 5b (CA5b) in pancreatic β cells. We furthermore discovered that pancreatic β cells express only one functional carbonic anhydrase isoform, CA5b, which is critical in replenishing oxaloacetate in the mitochondrial tricarboxylic acid (TCA) cycle (anaplerosis). These findings explain the mechanism for thiazide-induced glucose intolerance and reveal a fundamental role of CA5b in TCA cycle anaplerosis and insulin secretion in β cells.

Background

Thiazide diuretics are associated with glucose intolerance and new-onset diabetes mellitus. Previous studies demonstrated that thiazides attenuate insulin secretion, but the molecular mechanisms remain elusive. We hypothesized that thiazides attenuate insulin secretion via one of the known molecular thiazide targets in β cells.

Methods

We performed static insulin secretion experiments with islets of wild-type, Sodium/chloride co-transporter (NCC) (SLC12A3), and sodium-driven chloride/bicarbonate exchanger (NDCBE) (SLC4A8) knock-out (KO) mice and with murine Min6 cells with individual knockdown of carbonic anhydrase (CA) isoforms to identify the molecular target of thiazides in β cells. CA isoform 5b (CA5b) KO mice were then used to assess the role of the putative thiazide target CA5b in β-cell function and in mediating thiazide sensitivity in vitro and in vivo.

Results

Thiazides inhibited glucose- and sulfonylurea-stimulated insulin secretion in islets and Min6 cells at pharmacologically relevant concentrations. Inhibition of insulin secretion by thiazides was CO₂/HCO₃⁻-dependent, not additive to unselective CA inhibition with acetazolamide, and independent of extracellular potassium. By contrast, insulin secretion was unaltered in islets of mice lacking the known molecular thiazide targets NCC or NDCBE. CA expression profiling with subsequent knockdown of individual CA isoforms suggested mitochondrial CA5b as a molecular target. In support of these findings, thiazides significantly attenuated Krebs cycle anaplerosis through reduction of mitochondrial oxaloacetate synthesis. CA5b KO mice were resistant to thiazide-induced glucose intolerance, and thiazides did not alter insulin secretion in CA5b KO islets.

Conclusions

Thiazides attenuate insulin secretion via inhibition of the mitochondrial CA5b isoform in β cells of mice.

Introduction

Thiazide and thiazide-like diuretics (thiazides) have been the cornerstone for the treatment of essential hypertension and pharmacologic recurrence prevention of kidney stones for more than 50 years. Hence, not surprisingly, thiazides belong to the most widely prescribed drugs worldwide.¹

Since their introduction into clinical medicine in the 1960s, thiazides are known to be associated with metabolic side effects, including glucose intolerance and new-onset diabetes.^2–8 Several hypotheses have been put forth to explain thiazide-induced glucose intolerance, but the underlying mechanisms remain elusive until today.⁹ Unfortunately, these unpredictable and poorly understood side effects have caused many physicians to avoid the use of these clinically effective, ubiquitously available and cheap drugs. In recognition of this important knowledge gap in a clinically highly relevant area, a working group of the National Heart, Lung and Blood Institute issued a call for research on thiazide-induced dysglycemias a decade ago.⁹

The classical molecular thiazide target is the Na⁺/Cl⁻ cotransporter NCC (also known as SLC12A3) in distal convoluted tubules (DCTs) of the kidney.¹⁰ Biallelic pathogenic variants in SLC12A3 encoding NCC result in Gitelman syndrome which is characterized by hypotension, hypokalemia, hypomagnesemia, and metabolic alkalosis.¹¹ In addition to electrolyte abnormalities, patients affected by Gitelman syndrome were reported to exhibit an increased prevalence of impaired glucose tolerance.^12–14 Other molecular thiazide targets than NCC have been described, including the Na⁺-driven Cl⁻/bicarbonate exchanger NDCBE (also known as SLC4A8) and carbonic anhydrase (CA).^15,16 Thiazides were originally developed by chemical modification of the CA inhibitor acetazolamide (AZM) and retained the ability to inhibit CA.¹⁷ Previous studies demonstrated that AZM or thiazides at high doses attenuate insulin secretion in vitro, suggesting that inhibition of CA may play a role in thiazide-induced glucose intolerance.^18–20 Whether NCC or NDCBE are expressed in pancreatic islets and contribute to insulin secretion has not been explored thus far. The goal of this study was to elucidate the molecular mechanisms underlying thiazide-induced glucose intolerance. We hypothesized that thiazides attenuate insulin secretion in β cells via a known molecular thiazide target, likely a specific CA isoform.

Methods

Intraperitoneal Glucose and Insulin Tolerance Tests

Tolerance tests were performed in 10–12 week-old male mice after a 6 am to 12 pm 6-hour fast (intraperitoneal glucose tolerance test [IPGTT]) or at random fed state at 2 pm (intraperitoneal insulin tolerance test [IPITT]), as described.^21,22 Blood glucose was measured at time−30, 0, 15, 30, 60, and 120 minutes with a Contour glucose monitor (Bayer Healthcare, Germany) by tail vein sampling in duplicates. Vehicle or hydrochlorothiazide (HCT) was applied by intraperitoneal injection at time −30 minutes. Glucose (IPGTT; 1 or 2 g/kg, Sigma-Aldrich) or insulin (IPITT; 0.5 or 1 U/kg Actrapid HM, Novo Nordisk, Denmark) were applied by intraperitoneal injection at time 0 minute. For serum insulin measurements, tail vein blood sampling was performed at time −30, 0, and 2 minutes. Vehicle or HCT was applied at time −30 minutes and glucose (2 g/kg) was applied at time 0 minute. Serum insulin was measured with the ultrasensitive mouse insulin ELISA (CrystalChem, Downers Grove, IL; #90080).

Isolation of Islets and In Vitro Insulin Secretion Assays

Pancreata were perfused in situ with collagenase solution, and islets were isolated exactly as described.^22,23 After overnight incubation in Roswell Park Memorial Institute cell culture medium with 11 mM glucose, islets were washed twice with Krebs-Ringer bicarbonate buffer containing (in mM) 115 NaCl, 5 potassium chloride (KCI), 25 NaHCO₃, 0.5 NaH₂PO₄, 1.2 MgSO₄, 2.5 CaCl₂, 2 glucose, 10 N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid pH 7.4, and 0.1% (w:v) BSA. In case of experiments without CO₂/HCO₃⁻, the incubation buffer contained (in mM) 140 NaCl, 5 KCl, 0 NaHCO3, 0.5 NaH2PO4, 1.2 MgSO₄, 2.5 CaCl2, 2 glucose, 10 N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid pH 7.4, and 0.1% (w:v) BSA. After washing, islets were placed in 12-well plates (10 islets/well) containing 1 ml of buffer with 2 mM glucose and preincubated for 2.5 hours at 37°C. Insulin secretion into the supernatant was then measured for 2 hours for all secretagogues except for KCl, which was for 1 hour. Supernatants were then harvested, plates were put on ice, and total cellular insulin was extracted by addition of acid ethanol (70% ethanol, 1.5% HCl conc). Secreted and cellular insulin were determined with the ultrasensitive mouse insulin ELISA.

Statistical Analyses

Data distribution was assessed by D'Agostino-Pearson tests and QQ-plots. In case of deviation from a Gaussian distribution, nonparametric tests were used, and data were displayed as median±interquartile range. Data analysis was done with GraphPad Prism 9.3.1 (GraphPad Software, San Diego, CA). All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

Results

HCT Induces Glucose Intolerance in Mice Without Affecting Insulin Sensitivity

We first assessed if thiazides induce glucose intolerance in mice. To this end, IPGTTs were performed in three-month-old male C57BL/6J mice treated with intraperitoneally injected 1, 2, 5, or 50 mg/kg body weight HCT or vehicle. As shown in Figure 1, A–H, HCT-treated mice exhibited significantly higher glycemic excursions during IPGTTs compared with vehicle-treated mice in a dose-dependent manner. The lowest HCT dose associated with glucose intolerance was 5 mg/kg (Figure 1, E and F). In all conditions, body weight was similar in HCT-treated and vehicle-treated mice (Supplemental Figure 1, A–D). After overnight fasting, IPGTT results were similar compared with 6-hour fasted mice (Supplemental Figure 2, A–C). Plasma potassium levels were unaltered in HCT-treated compared with vehicle-treated mice during the IPGTT (Supplemental Figure 2D). Also with a lower glucose challenge (1 g/kg instead of 2 g/kg body weight), HCT-treated mice displayed significantly higher glycemia compared with vehicle-treated mice (Supplemental Figure 3). To assess the impact of HCT on insulin sensitivity, we performed IPITTs. As demonstrated in Figure 1, I–L and Supplemental Figure 1, E and F, insulin sensitivity, tested by two different insulin doses (0.5 and 1 IU/kg body weight, respectively), was not altered by HCT application. Together, these findings reveal that HCT induces glucose intolerance with maintained insulin sensitivity in mice.

Figure 1.

HCT induces glucose intolerance in mice without affecting insulin sensitivity. IPGTTs of C57BL/6J mice with corresponding area under the curve (AUC)s treated with 1 mg/kg (A and B), 2 mg/kg (C and D), 5 mg/kg (E and F), and 50 mg/kg (G and H) HCT or vehicle in 6-hour fasted mice. HCT or vehicle i.p. (arrow H) was applied at time point −30 minutes and glucose (2 g/kg; arrow G) was applied at time point 0 minute. Whole blood glucose was measured at indicated time points. IPITTs of C57BL/6J mice with corresponding AUCs treated with 5 mg/kg HCT or vehicle i.p. (arrow H) at time point −30 minutes and insulin (arrow I) at time point 0 minute. (I and J) Insulin 0.5 IU/kg i.p. and (K and L) insulin 1 IU/kg i.p. (A–H) N=10 mice per group in all experiments. Data are shown as mean±SD (A, C, E, G, I, and K) or median±interquartile range (IQR) (B, D, F, H, J, and L). Asterisks denote significance for the indicated comparisons (two-tailed unpaired Student t test (A, C, E, G, I, and K) and Mann-Whitney test (B, D, F, H, J, and L); *P < 0.05,**P < 0.01, ***P < 0.001, and ****P < 0.0001).

In humans, robust diuretic effects are observed with HCT doses of 1–2 mg/kg body weight.²⁴ Significantly higher thiazide doses are needed in mice to stimulate natriuresis; in case of HCT, typically doses of 20–50 mg/kg body weight are applied.^25–28 The reason for this discrepancy is not clear, but is likely due to differences in the pharmacokinetics of thiazides between mice and humans. To this end, we established the dose-response for the natriuretic effect of HCT in mice, which reflects inhibition of the primary thiazide target NCC in the DCT. As shown in Supplemental Figure 4, there was a dose-dependent increase in urinary Na⁺ excretion. The 5 mg/kg but not 1 mg/kg HCT dose increased urinary Na⁺ excretion compared with vehicle-treated mice, and the natriuretic effect was higher with 50 mg/kg compared with 5 mg/kg. Hence, the 5 mg/kg HCT dose in mice corresponds to a HCT dose typically employed in humans (12.5–50 mg daily with corresponding steady-state plasma concentrations of 0.02–0.2 µM).^29–31

Thiazides Attenuate Insulin Secretion In Vivo and In Vitro

To further investigate the basis of HCT-induced glucose intolerance, we measured serum insulin levels in mice treated with HCT or vehicle. As depicted in Figure 2, A and B and Supplemental Figure 1, G and H, serum insulin levels were significantly lower 2 minutes after an IPGTT challenge (2 g/kg body weight) in mice pretreated with either 5 or 50 mg/kg body weight HCT, indicating reduced first-phase insulin secretion. This notion is supported by the observation that HCT-treated mice exhibit hyperglycemia early during the IPGTT, followed by a decrease in blood glucose that parallels the decrease observed in vehicle-treated mice.

Figure 2.

HCT attenuates insulin secretion in vivo and in vitro. Serum insulin of C57BL/6J mice treated with 5 mg/kg (A) or 50 mg/kg (B) HCT or vehicle i.p. (arrow H) at time point −30 minutes and glucose (2 g/kg; arrow G) at time point 0 minute. Data are shown as median±IQR. (A and B) Asterisks denote significance for comparisons between groups of mice at indicated time points (Mann-Whitney test; ***P < 0.001). (C) glucose-stimulated insulin secretion (GSIS) (20 mM glucose) of islets isolated from C57BL/6J mice incubated with vehicle (DMSO 1:1000) or HCT at indicated concentrations. Each dot represents islets isolated from an individual mouse. (D) GSIS of Min6 cells incubated with vehicle or HCT at indicated concentrations. Data represent three individual experiments combined. (E) Insulin secretion of Min6 cells in the presence of indicated concentrations of glucose (2 or 20 mM) and potassium chloride (KCI) (5 or 30 mM), treated with 10⁻⁸ M HCT or 250 µM tolbutamide. Data represent three individual experiments combined. (C–E) Data are shown as individual observations with mean±SD. Asterisks denote significance for the indicated comparisons (ANOVA with Tukey post hoc test; **P < 0.01 and ****P < 0.0001).

We then assessed the impact of HCT on insulin secretion in primary islets isolated from C57BL/6J mice and in the murine β-cell line Min6 (Figure 2, C and D, Supplemental Figure 5). HCT significantly attenuated insulin secretion in islets and Min6 cells in vitro at pharmacologically relevant concentrations while not affecting cell viability.³¹ We obtained similar findings with other thiazides, including metolazone, chlorthalidone, indapamide, and bendroflumethiazide (Supplemental Figure 6). HCT also significantly lowered sulfonylurea-induced insulin secretion (250 µM tolbutamide) while it only had a small effect on basal (2 mM glucose) and direct depolarization-induced insulin secretion by high (30 mM) extracellular K⁺ (Figure 2E).³² Together, these results suggest reduced insulin secretion as a mechanism of HCT-induced glucose intolerance.

Thiazides Target CA in β Cells

To further define the basis of reduced insulin secretion by thiazides, we assessed mRNA expression of the known thiazide targets, including NCC, NDCBE, and CA isoforms in Min6 cells and murine islets (Supplemental Figure 7, A–D). Transcripts of both NDCBE and NCC were detectable in Min6 cells and islets. Of the 15 CA isoforms CA1, CA4, and CA5a were not detectable in both Min6 cells and islets. CA6 and CA7 were expressed in Min6 cells but not islets. We next assed the impact of genetic NDCBE or NCC deletion in islets on insulin secretion. As shown in Supplemental Figure 7, E and F, islets isolated of NDCBE and NCC knock-out (KO) mice displayed no insulin secretion deficit. To assess the role of NCC in systemic glucose homeostasis, we additionally performed IPGTTs in NCC KO mice. As shown in Supplemental Figure 7, G and H, genetic loss of NCC was not associated with altered glucose tolerance. Furthermore, HCT attenuated insulin secretion to a similar degree in islets of NCC KO mice as in islets of wild-type (WT) mice (Supplemental Figure 7I).

We then assessed the role of CA in insulin secretion. Inhibition of CA with the nonspecific CA inhibitor AZM significantly attenuated insulin secretion in both Min6 cells and islets in vitro (Figure 3, A and B) and induced glucose intolerance in vivo (Figure 3, C–E). Compared with either AZM or HCT alone, the combination of AZM and HCT did not further attenuate insulin secretion (Figure 3F), and HCT had no effect on insulin secretion in a CO₂/HCO₃⁻-free condition (Figure 3G). These results reveal that the effect of HCT on insulin secretion is CO₂/HCO₃⁻-dependent and not additive to (unselective) CA inhibition, suggesting that HCT targets one or several CA isoform(s) in β cells that are critical for insulin secretion.

Figure 3.

HCT targets CA in β ells. (A) GSIS (20 mM glucose) of Min6 cells incubated with vehicle (DMSO 1:1000) or the CA inhibitor AZM at indicated concentrations. (B) GSIS of islets isolated from C57BL/6J mice incubated with vehicle or AZM at indicated concentrations. Each dot represents islets isolated of an individual mouse. (C–E) IPGTTs with corresponding AUCs and body weights of C57BL/6J mice treated with 5 mg/kg AZM. AZM or vehicle i.p. (arrow A) was applied at time point—30 minutes and glucose (2 g/kg; arrow G) was applied at time point 0 minute. Whole blood glucose was measured at indicated time points. N=8 mice per group. (F) GSIS of Min6 cells treated with vehicle or indicated concentrations (M) of either AZM or HCT or AZM and HCT combined. (G) GSIS of Min6 cells in the presence (+) or absence (−) of CO₂/HCO₃⁻, treated with vehicle or HCT at indicated concentrations. (H) GSIS of Min6 cells treated with control small interfering RNA (siRNA) or siRNAs targeting indicated CA isoforms. (I) GSIS of Min6 cells treated with control siRNA or siRNAs targeting CA2 or CA5b, respectively, in the presence (+) or absence (−) of CO₂/HCO₃⁻. (J) Expression profiling of CA isoforms in purified murine β cells by real-time PCR and normalized to glycerinaldehyd-3-phosphat-dehydrogenase. Each dot represents β cells isolated from an individual mouse. (K) Assessment of purity of isolated islets and purified β cells by transcript expression analysis of amylase (marker of exocrine pancreas), glucagon (marker of α cells), and insulin (marker of β cells) in pancreas, islets, and purified β cells by real-time PCR and normalized to glycerinaldehyd-3-phosphat-dehydrogenase. Each dot represents tissue or cells isolated from an individual mouse. Min6 cell experiments represent three individual experiments combined. Data are shown as mean±SD except in (D and F) where median±IQR is displayed. Asterisks denote significance for the indicated comparisons (ANOVA with Tukey post hoc test panels (A, B, and I); two-tailed unpaired Student t test panel (C); Mann-Whitney test (D); Kruskal-Wallis with Dunn post hoc test (F); *P < 0.05, **P < 0.01, ***P < 0.001; and ****P < 0.0001).

Currently, there are no specific CA isoform inhibitors available. To identify the responsible CA isoform(s), we performed small interfering RNA (siRNA)-mediated knockdown of all CA isoforms expressed in Min6 cells (Figure 3H, Supplemental Figures 7C and 8) and performed static insulin secretion experiments. Only CA2 and CA isoform 5b (CA5b) knockdown significantly attenuated insulin secretion in Min6 cells, whereas the other CA isoforms expressed in Min6 cells were dispensable for insulin secretion (Figure 3H). In addition, omission of CO₂/HCO₃⁻ in the incubation medium did not further impair insulin secretion in either CA2- or CA5b-depleted cells compared with control, indicating critical but nonredundant roles of these two CA isoforms for insulin secretion in Min6 cells (Figure 3I). We then assessed CA isoform mRNA expression in murine islets and purified β cells. As shown in Figure 3, J and K and Supplemental Figure 7D, although many CA isoforms are expressed in islets, primary β cells only express the two isoforms CA5b and CA10. These findings, together with the fact that CA10 is a catalytically inactive and secreted CA isoform, suggest mitochondrial CA5b as the likely target of thiazides in β cells.³³

HCT Inhibits Oxaloacetate Synthesis in β Cells

The two mitochondrial CA isoforms CA5a and CA5b use CO₂, which freely diffuses into mitochondria, to produce HCO₃⁻ (Supplemental Figure 11). Several mitochondrial enzymes critically depend on HCO₃⁻, such as pyruvate carboxylase (PC), which generates oxaloacetate (OAA) from pyruvate and HCO₃⁻³⁴ β cells exhibit high PC activity, and a large fraction of pyruvate entering mitochondria is converted to OAA. PC activity correlates steeply with insulin secretion and OAA is a central metabolite in nutrient-induced insulin secretion (Supplemental Figure 11).³⁵ OAA synthesis by PC (anaplerosis) fuels the tricarboxylic acid (TCA) cycle leading to an increase of TCA intermediates, such as citrate, isocitrate, and malate, that are transported from the mitochondria to the cytoplasm (cataplerosis). OAA is also the starting point of phosphoenolpyruvate (PEP) synthesis via the mitochondrial guanosine-5′-triphosphate-dependent enzyme PCK2.^35–38 PEP synthesized from OAA exits mitochondria to the cytosol, where pyruvate kinase converts ADP and PEP into ATP and pyruvate, leading to closure of K_ATP channels and initiation of insulin secretion.^39,40 The third OAA-dependent pathway contributing to nutrient-stimulated insulin secretion is the pyruvate/malate shuttle, which results in the generation of cytosolic NADPH via malic enzyme.^35,36

In a next step, we measured glucose-induced insulin secretion in Min6 cells in the presence of HCT, the PC inhibitor phenylacetic acid (PAA) or a combination of both, with and without CO₂/HCO₃⁻. As shown in Figure 4A, inhibition of glucose-stimulated insulin secretion (GSIS) was equal with incubation of HCT or PAA, and the combination of both did not result in a further reduction. Furthermore, HCT, PAA, or the combination of both had no effect on GSIS in CO₂/HCO₃⁻-free conditions. Similarly, in murine islets, GSIS was reduced to a similar degree in the presence of HCT or PAA, and the combination of both HCT and PAA did not result in a further reduction (Figure 4B).

Figure 4.

HCT attenuates OAA synthesis in β cells. (A) GSIS (20 mM glucose) of Min6 cells in the presence (+) or absence (−) of CO₂/HCO₃⁻, treated with vehicle (DMSO 1:1000), HCT, the PC inhibitor PAA, or a combination of PAA and HCT at indicated concentrations. (B) GSIS of islets isolated from C57BL/6J mice incubated vehicle, HCT, PAA, or a combination of PAA and HCT at indicated concentrations. (C) OAA content normalized to protein of Min6 cells stimulated with 20 mM glucose, in the presence (+) or absence (−) of CO₂/HCO₃⁻ and treated with either vehicle, HCT, PAA, or a combination of HCT and PAA at indicated concentrations. (D) OAA content normalized to protein of Min6 cells stimulated with 20 mM glucose, in the presence (+) or absence (−) of CO₂/HCO₃⁻, treated with control siRNA or siRNA targeting CA5b, and incubated with either vehicle or HCT at indicated concentrations. (E) PC activity in lysates of Min6 cells cultured for 4 hours in the absence of CO₂ before lysis, measured in CO₂-free condition in the presence of indicated concentrations of HCT or AZM. Data are shown as individual observations with mean±SD and represent three individual experiments combined. Asterisks denote significance for the indicated comparisons (ANOVA with Tukey post hoc test; ****P < 0.0001).

We then measured OAA levels in Min6 cells exposed to HCT or PAA during glucose stimulation. As shown in Figure 4C, both HCT and PAA attenuated OAA levels in Min6 cells by a similar magnitude, and the combination of both did not result in a further reduction of OAA levels. Furthermore, HCT, PAA, or the combination of both had no effect on OAA levels in Min6 cells in CO₂/HCO₃⁻-free conditions. Finally, we treated Min6 cells with control or CA5b siRNA and measured OAA in the presence and absence of CO₂/HCO₃⁻. CA5b depletion or HCT were equally effective in reducing OAA levels in Min6 cells, and this effect was again clearly CO₂/HCO₃⁻-dependent (Figure 4D).

Because of the tight functional coupling between CA5b and PC, a direct effect of thiazides on PC activity cannot be definitively ruled out with studies using intact cells. To test for a possible effect of HCT on PC, we performed in vitro PC activity experiments in Min6 cell lysates in the absence of CO₂ with exogenous administration of HCO₃⁻. As demonstrated in Figure 4E, HCT, AZM, or the combination of HCT and AZM had no effect on PC activity, further supporting the notion of CA5b as molecular target of HCT in β cells. To substantiate this claim, we performed IPGTTs in CA5b KO and WT littermate mice (mixed C57BL6/J/SV129 background) treated with 5 mg/kg body weight HCT or vehicle.³⁴ HCT-treated WT mice exhibited significantly higher glycemic excursions during IPGTTs compared with vehicle-treated mice (Figure 5, A and B, Supplemental Figure 9A). By contrast, we observed no difference between HCT- and vehicle-treated CA5b KO mice (Figure 5, C and D, Supplemental Figure 9B). Acid-base parameters and electrolytes were similar in both groups of mice (Supplemental Table 1). We then performed IPITTs to assess the impact of HCT on insulin sensitivity in 3-month-old male WT and CA5b KO mice. As demonstrated in Figure 5, E–H and Supplemental Figure 9, C and D, insulin sensitivity was similar in all groups of mice. Furthermore, in WT but not CA5b KO mice, HCT treatment was associated with lower serum insulin after the glucose challenge compared with vehicle treatment (Figure 5, I and J, Supplemental Figure 9, E and F). Together, these in vivo findings demonstrate that genetic deletion of CA5b confers resistance to HCT-induced glucose intolerance.

Figure 5.

CA5b KO mice are resistant to HCT-induced glucose intolerance. IPGTTs in WT (A and B) and CA5b KO (C and D) mice with corresponding AUCs treated with 5 mg/kg HCT. HCT or vehicle i.p. (arrow H) was applied at time point −30 minutes and glucose (2 g/kg; arrow G) was applied at time point 0 minute. Whole blood glucose was measured at indicated time points. IPITTs in WT (E and F) and CA5b KO (G and H) mice with corresponding AUCs treated with 5 mg/kg HCT or vehicle i.p. (arrow H) at time point −30 minutes and insulin (1 IU/kg i.p.; arrow I) at time point 0 minute. Serum insulin in WT (I) and CA5b KO (J) mice treated with 5 mg/kg HCT or vehicle i.p. (arrow H) at time point −30 minutes and insulin (2 g/kg; arrow G) at time point 0 minute. N=10 mice per group in all experiments. Data are show as mean±SD. Asterisks denote significance for the indicated comparisons (two-tailed unpaired Student t test; ****P < 0.0001).

To assess the role of genetic CA5b deletion on insulin secretion, we performed static insulin secretion experiments with islets isolated from WT and CA5b KO mice (Figure 6A). No CA5b expression was detectable in islets isolated from CA5b KO mice (Supplemental Figure 10A). Furthermore, expression profiling of other CA isoforms did not reveal differences between WT and CA5b KO islets (Supplemental Figure 10B). As previously observed with islets isolated from C57BL/6J mice, HCT attenuated insulin secretion in islets isolated from WT littermates of CA5b KO mice (mixed C57BL/6J/SV129 background). Although insulin secretion of CA5b KO islets was significantly reduced compared with WT islets in the presence of the vehicle, HCT had no effect on insulin secretion of CA5b KO islets. A similar pattern was observed when we quantified OAA levels in WT and CA5b KO islets (Figure 6B). In WT islets, HCT attenuated both high glucose (20 mM)– and sulfonylurea (250 µM tolbutamide)–induced insulin secretion but did not influence basal (2 mM glucose) or direct depolarization-induced insulin secretion by high (30 mM) extracellular K⁺ (Figure 6C). By contrast, HCT did not attenuate basal or stimulated insulin secretion in CA5b KO islets (Figure 6D). Together these results demonstrate that mitochondrial OAA synthesis and insulin secretion capacity of CA5b KO islets are resistant to the action of HCT.

Figure 6.

HCT does not attenuate insulin secretion and OAA levels in CA5b KO islets. (A) GSIS (20 mM glucose) of WT (filled circles) and CA5b KO (open circles) islets treated with vehicle (DMSO 1:1000) or HCT at indicated concentrations. (B) OAA content normalized to protein of WT (filled circles) and CA5b KO (open circles) islets stimulated with 20 mM glucose and incubated with vehicle or HCT at indicated concentrations. Insulin secretion of WT (C) or CA5b KO (D) islets in the presence of indicated concentrations of glucose (2 or 20 mM) and KCl (5 or 30 mM), incubated with vehicle, HCT (10⁻⁸ M), or tolbutamide (250 µM). Data are shown as individual observations with mean±SD. Each dot represents islets isolated from an individual mouse. Asterisks denote significance for the indicated comparisons (ANOVA with Tukey post hoc test; ****P < 0.0001).

Discussion

Our study reveals that thiazides induce acute glucose intolerance in mice via attenuation of insulin secretion through inhibition of CA5b in β cells. HCT and other frequently used thiazides, such as chlorthalidone, indapamide, metolazone, and bendroflumethiazide, inhibited insulin secretion in a pharmacologically relevant, submicromolar range. CA expression profiling in Min6 cells and islets and subsequent siRNA knockdown experiments of individual CA isoforms in Min6 cells suggested the mitochondrial CA5b isoform as a molecular target of thiazides in β cells. In support of these results, CA5b KO mice were resistant to HCT-induced glucose intolerance, and insulin secretion of CA5b-deficient islets or Min6 cells was unaffected by HCT.

Mitochondrial CA5b provides HCO₃^- for anaplerotic OAA synthesis from pyruvate by PC (Supplemental Figure 11).³⁵ Deletion of CA5b or treatment with HCT greatly attenuated OAA levels in islets or Min6 cells. In line with these findings, purified full length (human) CA5b was previously shown to be directly inhibited by HCT and other thiazides in vitro with a K_i in the nanomolar range.¹⁷ Although direct PC inhibition by PAA in islets mimicked the findings obtained with HCT treatment or CA5b deletion, our in vitro studies in CO₂-free conditions with exogenous administration of HCO₃⁻ demonstrate that PC is not inhibited by HCT or other thiazides up to the concentration of 10⁻⁵ M. This experiment enabled to functionally separate the two closely interacting enzymes CA5b and PC and led to the conclusion that PC per se is not thiazide-sensitive. In support of these results, structural studies demonstrated direct interaction of thiazides with CA isoforms.^17,41

OAA is a well-established central metabolite in nutrient-induced insulin secretion.³⁵ In support of this, we found that secretagogue-induced insulin secretion was severely impaired in CA5b KO islets. Attenuation of insulin secretion was more pronounced in CA5b KO islets compared with acute knockdown with siRNA or treatment with thiazides, primarily due to significantly increased basal insulin secretion in CA5b KO islets. CA5b KO islets were completely unresponsive to stimulation by high glucose, tolbutamide, and high extracellular K⁺.

In Min6 cells but not WT islets, HCT also slightly reduced basal and high K⁺-stimulated insulin secretion. The reason for these differences is not clear at the moment. We speculate that residual CA5b activity (inhibition, siRNA knockdown) versus complete loss of CA5b activity (KO model), and acute (inhibition, siRNA knockdown) versus chronic CA5b deficiency (KO model) may play a role.

Interestingly, Min6 cells express several CA isoforms and seem to depend on cytosolic CA2 activity in addition to mitochondrial CA5b for insulin secretion. By contrast, primary murine β cells only express CA5b and CA10. Although CA2-positive pancreatic cells are progenitors of both exocrine and endocrine pancreatic cells, CA2 remains highly expressed in the exocrine but not endocrine pancreas on differentiation.⁴² Similarly, immortalization procedures, clonal selection artifacts, and/or adaptations to cell culture conditions are responsible for the altered CA expression profile and the acquired dependence on CA2 for insulin secretion in Min6 cells. As is the case for CA5b, CA10 is also expressed in both Min6 cells and primary β cells. CA10 is a catalytically inactive, secreted glycoprotein that was recently shown to physically interact with neurexins, a family of presynaptic adhesion molecules, and to facilitate their surface transport.^33,43 Hence, our data intriguingly suggest that β cells do not express a cytoplasmic CA isoform. Furthermore, only one of the two mitochondrial CA isoforms is expressed in β cells. CA5b KO mice do not exhibit an overt phenotype.³⁴ By contrast, CA5a KO mice exhibited reduced growth, poor fertility, and hyperammonemia as a result of defective ureagenesis in the liver. This suggests nonredundant physiological roles of the two mitochondrial CA isoforms. Purified CA5a can also be inhibited by thiazides in vitro, but the K_i values for thiazides are much higher compared with CA5b.¹⁷ Nevertheless, an effect of thiazides on CA5a activity with therapeutic doses of thiazides would theoretically still be possible and should be investigated further.

In the past, several hypotheses have been proposed for thiazide-induced glucose intolerance, including decreased peripheral^44–46 or hepatic insulin sensitivity,^47,48 activation of the sympathetic nervous and the renin angiotensin system by thiazides,⁹ or attenuation of insulin secretion because of thiazide-induced hypokalemia.⁴⁹ In our acute model, insulin sensitivity was not affected by HCT and not different between WT and CA5b KO mice. In addition, we observed thiazide-induced attenuation of insulin secretion in the absence of changes in extracellular K⁺.

Our study also has limitations, such as the lack of a conditional CA5b KO mouse model to exclude the possibility that pathways outside the β cells contribute to thiazide-induced glucose intolerance. In addition, although insulin tolerance tests did not display differences between vehicle- and thiazide-treated groups, we did not perform hyperinsulinemic euglycemic clamping experiments to definitively rule out an impact of HCT on hepatic gluconeogenesis. Furthermore, we did not assess the impact of chronic thiazide administration on glucose metabolism. It is possible that additional mechanisms contribute to thiazide-induced glucose intolerance during long-term administration of thiazides. Clearly, however, acute thiazide administration results in CA5b-dependent attenuation of insulin secretion with subsequent glucose intolerance. Furthermore, chronic mitochondrial CA5b deficiency in β cells cannot be fully compensated. Although the basal insulin secretion is increased, islets of mice with a constitutive CA5b deletion secrete significantly less insulin when stimulated with glucose or sulfonylureas compared with islets of WT mice.

In summary, our results demonstrate that thiazides induce glucose intolerance by an attenuation of insulin secretion in β cells through inhibition of mitochondrial CA5b.

Supplementary Material

jasn-34-1179-s001.pdf ^{(1.1MB, pdf)}

Disclosures

M. Anderegg reports Ownership Interest: small numbers: Alnylam Pharmaceuticals, Bayer AG, BioNTech, Moderna, Novartis, and Roche. D.G. Fuster reports Consultancy: Alnylam Pharmaceuticals, Kyowa Kirin, and Otsuka Pharmaceuticals Switzerland; Research Funding: Boehringer Ingelheim and Otsuka Pharmaceuticals Switzerland; Honoraria: Alnylam Pharmaceuticals and Otsuka Pharmaceuticals Switzerland; and Advisory or Leadership Role: Alnylam Pharmaceuticals, Kyowa Kirin, and Otsuka Pharmaceuticals Switzerland. D. Konrad reports Research Funding: NovoNordisk. The remaining authors have nothing to disclose.

Funding

D.G. Fuster was supported by the Swiss National Science Foundation (#Grant 31003A_172 974), the Swiss National Centers of Competence in Research (NCCR TransCure and NCCR Kidney.CH), the Novartis Research Foundation, and by a Medical Research Position Award of the Foundation Prof. Dr. Max Cloëtta.

Author Contributions

Conceptualization: Daniel G. Fuster.

Data curation: Giuseppe Albano, Christine Deisl, Daniel G. Fuster, Patrycja Kucharczyk.

Formal analysis: Giuseppe Albano, Matteo Bargagli, Daniel G. Fuster, Daniel Konrad, Patrycja Kucharczyk, Stephan Wueest.

Funding acquisition: Daniel G. Fuster, Daniel Konrad.

Investigation: Giuseppe Albano, Manuel Anderegg, Matteo Bargagli, Christine Deisl, Daniel G. Fuster, Tin Manh Ho, Patrycja Kucharczyk, Stephan Wueest.

Methodology: Giuseppe Albano, Christine Deisl, Daniel G. Fuster, Daniel Konrad, Patrycja Kucharczyk, Stephan Wueest.

Project administration: Giuseppe Albano, Daniel G. Fuster.

Resources: Daniel G. Fuster.

Supervision: Daniel G. Fuster, Daniel Konrad, Stephan Wueest.

Writing – original draft: Daniel G. Fuster.

Writing – review & editing: Giuseppe Albano, Manuel Anderegg, Matteo Bargagli, Christine Deisl, Daniel G. Fuster, Tin Manh Ho, Daniel Konrad, Patrycja Kucharczyk, Stephan Wueest.

Data Sharing Statement

All raw data, processed data, and corresponding metadata will be made available on Dryad (https://doi.org/10.5061/dryad.5qfttdz9v) on publication of the manuscript.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/JSN/E387.

Supplemental Methods.

Supplemental Figure 1. Body weights of mice of in vivo experiments displayed in Figure 1.

Supplemental Figure 2. Glucose tolerance test with hydrochlorothiazide treatment after overnight fasting.

Supplemental Figure 3. Glucose tolerance test with hydrochlorothiazide treatment with reduced dose (1 g/kg) glucose challenge.

Supplemental Figure 4. Dose dependence of the natriuretic effect of hydrochlorothiazide in mice.

Supplemental Figure 5. Assessment of cell viability in murine islets or Min6 cells treated with hydrochlorothiazide.

Supplemental Figure 6. Effect of thiazide diuretics and thiazide-like diuretics on insulin secretion.

Supplemental Figure 7. Expression of thiazide targets in Min6 cells and islets and insulin secretion studies of islets isolated from NCC and NDCBE KO mice.

Supplemental Figure 8. Knockdown of CA isoforms in Min6 cells by siRNA.

Supplemental Figure 9. Body weights of mice of in vivo experiments displayed in Figure 4.

Supplemental Figure 10. CA isoform expression in islets of WT and CA5b KO mice.

Supplemental Figure 11. Schematic of a islet β cell.

Supplemental Table 1. Acid-base parameters and electrolytes of WT and CA5b KO mice.