Sirolimus (rapamycin) is an immunosuppressive drug used in transplantation. One of its major side effects is the increased risk of diabetes mellitus; however, the exact mechanisms underlying such association have not been elucidated. Here we show that sirolimus impairs glucose-stimulated insulin secretion both in human and murine pancreatic islets and in clonal β cells in a dose- and time-dependent manner. Importantly, we demonstrate that sirolimus markedly depletes calcium (Ca2+) content in the endoplasmic reticulum and significantly decreases glucose-stimulated mitochondrial Ca2+ uptake. Crucially, the reduced mitochondrial Ca2+ uptake is mirrored by a significant impairment in mitochondrial respiration. Taken together, our findings indicate that sirolimus causes depletion of intracellular Ca2+ stores and alters mitochondrial fitness, eventually leading to decreased insulin release. Our results provide a novel molecular mechanism underlying the increased incidence of diabetes mellitus in patients treated with this drug.
Post-transplant diabetes mellitus represents a major adverse effect of immunosuppressive drugs1,2,3,4 and is associated with high cumulative incidence of cardiac events, vascular disease, and overall impaired survival rates5. Sirolimus (rapamycin) was introduced in the Edmonton immunosuppression protocol in islet transplant recipients6,7, attempting to minimize the diabetogenic effects observed with corticosteroids and other immunosuppressive regimens. Despite the initial enthusiasm, 5-year results of this clinical trial revealed that only ~10% of patients maintained insulin independence1,8, endorsing the detrimental role of sirolimus in glucose homeostasis.
A randomized trial of immunosuppressive drugs in kidney transplantation, the Efficacy Limiting Toxicity Elimination (ELITE) – Symphony study9, identified sirolimus as the one with the highest incidence of hyperglycemia, even higher than calcineurin inhibitors9. Since then, several investigators sought to determine the mechanisms underlying new-onset diabetes mellitus after transplantation10,11,12,13. The effects of sirolimus in vivo are quite complex, as confirmed by numerous controversial findings: indeed, albeit several studies demonstrate that its administration causes glucose intolerance14,15,16, there are also reports showing that it does improve insulin sensitivity in diabetic mice17, protects against obesity18,19, reduces atherosclerosis20,21 and cardiac or renal fibrosis22,23, and extends lifespan24.
We decided to test the effects of sirolimus in pancreatic β cells. Our hypothesis is that one of the mechanisms underlying the diabetogenic action of sirolimus is the impairment of metabolism-secretion coupling in β cells. We focused on the effect of sirolimus on the key organelle in metabolism-secretion coupling, i.e. the mitochondrion25,26,27,28. Indeed, such organelle is considered the main responsible for coupling different fuel secretagogues to insulin exocytosis, through a process that includes oxidation of nutrients within the mitochondrial matrix and subsequent ATP generation, increasing intracellular calcium (Ca2+) via closure of ATP-sensitive K+ channels and depolarization of the plasma membrane27,28,29,30,31,32.
Sirolimus impairs glucose-induced insulin secretion in pancreatic β cells
To test the effect of sirolimus on pancreatic β cell function, we evaluated the response to glucose in INS-1 β cells. We first performed a dose-response assay, and we found that increasing doses of sirolimus progressively reduce glucose-stimulated insulin secretion (GSIS, Fig. 1a). Then, we performed a time-course experiment using the dose of sirolimus (25 nM) that has been measured in the blood of transplant recipients33 and we observed that a 24-hour incubation significantly decreased GSIS (Fig. 1b). Importantly, we did not detect any significant effect of sirolimus on cell viability (Fig. 1c). These results were also confirmed in murine (Fig. 1d,e) and in human (Fig. 1f,g) islets.
Sirolimus reduces mitochondrial respiration in pancreatic β cells
When testing the effect of sirolimus on insulin release in response to the fuel secretagogues leucine and glutamine, which are known to stimulate insulin exocytosis through increased mitochondrial metabolism34, we found a significantly impaired response in sirolimus-treated cells (Fig. 2a), whereas cells from both groups were similarly responsive to KCl-mediated depolarization (Fig. 2b). We obtained similar findings in murine (Fig. 2c,d) and in human (Fig. 1e,f) islets, thereby suggesting an action of the immunosuppressant drug on the mechanisms underlying metabolism-secretion coupling. Therefore, we tested the effect of sirolimus on mitochondrial respiration, observing a significant decrease in oxygen consumption rate (OCR) in clonal β cells (Fig. 3) and in islets isolated from mice and humans (Figure S1) treated with sirolimus.
Sirolimus decreases mitochondrial Ca2+ uptake in β cells
Mounting evidence indicates that Ca2+ represents a major regulator of mitochondrial function, and a decreased uptake of this ion by this organelle has been functionally linked to reduced mitochondrial respiration in various cell types35,36,37,38,39. Thus, we assessed mitochondrial Ca2+ uptake in pancreatic β cells following incubation with sirolimus and we observed a significantly decreased uptake compared with vehicle-treated cells (Fig. 4a,b). We and others have demonstrated the importance of ER Ca2+ in β cell function31,40,41,42,43,44,45. Since sirolimus has been reported to modulate intracellular Ca2+ fluxes in different tissues46,47,48,49, we assessed ER Ca2+ stores and we found that sirolimus-treated β cells exhibited depleted intracellular Ca2+ stores and increased Ca2+ leak (Fig. 4c,d). When measuring cytosolic Ca2+ levels, we also observed a slightly reduced response to glucose in sirolimus-treated β cells (Figure S2), further supporting our data on decreased GSIS.
Sirolimus regulates the expression of inositol 1,4,5-trisphosphate receptor in clonal β cells and human and murine islets
Since sirolimus is known to modulate transcriptional activity50,51,52, we tested its effect on the expression levels of key players in Ca2+ handling, namely inositol 1,4,5-trisphosphate receptor (IP3R)53, ryanodine receptor (RyR)31,54 and sarco/endoplasmic reticulum Ca2+-ATPase (SERCA)44 in clonal β cells and human and murine islets. We observed a significant upregulation of all three IP3R isoforms in sirolimus-treated compared with vehicle-treated cells, a result that was consistent in all of the tested species (rat, mouse, human, Fig. 5), strongly suggesting that the modulation of IP3R is one of the mechanisms underlying the effects of sirolimus on pancreatic β cells.
Immunosuppressive therapy has been shown to be associated with glucose intolerance and post-transplantation diabetes mellitus; such a diabetogenic effect is common to immune-modulating agents acting via different mechanisms of action3,9,55, including β cell failure4,13,56,57 and alteration of metabolic parameters controlled by insulin signaling58,59,60,61. Here we show that sirolimus has a detrimental effect on Ca2+ handling and GSIS both in human and murine islets and in rat insulinoma cell line INS-1.
We also demonstrate that at 25 nM – the dose measured in the blood of transplanted patients33 – sirolimus does not significantly affect cell vitality, consistent with previous experiments performed in various cell types revealing that only supra-therapeutic doses of sirolimus lead to increased apoptosis and reduced cell proliferation49,56,62, without affecting overall insulin content56. Furthermore, earlier investigations had established that immunosuppressant drugs, when used at concentrations comparable with therapeutic levels in humans, do not cause apoptosis in pancreatic β cells63.
We show here for the first time that sirolimus significantly compromises mitochondrial respiration, directly assessed by measuring oxygen consumption, both in pancreatic islets and clonal β cells. The direct modulation of intracellular Ca2+ release channels on the ER offers a novel mechanistic insight on the diabetogenic effect of sirolimus. Intriguingly, the upregulation of IP3R observed in pancreatic β cells following sirolimus treatment is consistent with the recent observation that IP3R levels are increased in islets from diabetic patients, mirrored by a reduced number of interactions between ER and mitochondria64. Additionally, Madec and colleagues had shown that exposing pancreatic islets to high glucose concentrations led to increased levels of IP3R65. Also, mutations in the gene encoding for IP3R have been associated to perturbations in glucose homeostasis and enhanced susceptibility to diet-induced diabetes mellitus66. Further studies are necessary to better delineate the exact role of IP3Rs in the regulation of Ca2+ fluxes in β cells and to identify other potential mechanisms.
Interestingly, the properties of sirolimus observed in β cells are cell-specific and seem to be in contrast with its effects seen in models of neurodegenerative and ischemic disorders67,68, in which the drug has been shown to be overall protective, inducing autophagy and enhancing lysosomal activation in order to remove damaged mitochondria69. The complexity of the pathways induced by sirolimus is further confirmed by the experimental findings of Fuhrer and colleagues, who observed that, despite sirolimus significantly suppresses β cell response to glucose (in agreement with our findings), the incubation of RIN-5F cells with high doses of sirolimus in absence of glucose can instead increase insulin secretion57. However, opposite to their results, Barlow and colleagues found that 200 nM sirolimus caused a significant reduction in both basal and glucose-stimulated insulin release in Min-6 cells56. The exact mechanisms underlying such different pharmacologic responses need to be characterized in future studies.
Materials and Methods
Human islets with >90% purity and viability were obtained from non-diabetic de-identified cadaveric donors through the Integrated Islet Distribution Program (IIDP). The characteristics of the donors are reported in Supplementary Table 1. Upon receipt, the islets were cultured as described70. Murine islets of Langerhans were isolated as previously described31. Procedures on rodents have been performed according to guidelines and regulations approved by the Einstein Animal Care and Use Committee. INS-1 cells were maintained in monolayer culture in RPMI-1640 medium, as previously described by our group71. Insulin levels were determined as described and validated31,71,72,73. In some experiments the cells were treated with glucose (5.5 and 16.7 mM, Bio-Techne, Abingdon, UK), sirolimus (LC Laboratories, Woburn, MA, dissolved in dymethylsulfoxide), or L-leucine (10 mM, MyBioSource, San Diego, CA, USA) and glutamine (2 mM, MyBioSource). Cell viability was estimated by the [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide, MTT colorimetric assay, spectophotometrically (570 nm) measuring the ability of metabolically active cells to reduce MTT.
Extracellular flux analyses
Extracellular flux analyses were performed using the Extracellular Flux Analyzer (Agilent Technologies, Santa Clara, CA, USA), according to the manufacturer’s instructions. Specifically, the following drugs were added to each well: glucose (16.7 mM, at minute 32) to determine response to high glucose; oligomycin (1 μM, at minute 104) to inhibit ATP synthase and assess coupling efficiency; carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP, 0.5 μM, at minute 176) to uncouple the mitochondrial oxidative phosphorylation and measure both maximum respiration and spare capacity; antimycin A and rotenone (both 1 μM, at minute 216) to inhibit the respiratory chain and measure non-mitochondrial respiration. After each assay, the cells were collected to quantify DNA via QuantiFluor dsDNA System (Promega, Madison, WI, USA), according to the manufacturer’s instructions.
Ca2+ dynamics in cytosol, ER and mitochondria
Ca2+ dynamics were evaluated as previously described and validated74,75,76,77. Briefly, cells attached on glass bottom culture dishes (MatTek Corporation, Ashland, MA) were loaded with Fura-2 acetoxymethyl (AM) ester (Thermo Fisher Scientific, Waltham, MA, USA, 5 μM, 15 min, 37 °C). Images were obtained using a dual excitation fluorescence imaging system, as described31: changes in intracellular Ca2+ were expressed as the ratio of fluorescence emission acquired above 510 nm in response to excitation at 340 nm and 380 nm. ER Ca2+ was assessed in cells transfected with the luminal Ca2+ sensor D1ER (Addgene, Cambridge, MA), as described78,79,80 and the rate of Ca2+ leak was measured as function of [Ca2+]ER following the addition of thapsigargin (1 μM). To evaluate mitochondrial Ca2+, the samples were loaded with rhod-2 AM (Thermo Fisher Scientific, 3 μM, 30 min, 37 °C), followed by washout and 1 hour rest at room temperature for de-esterification31,77,81. Fluorescence was detected using a pass-band filter of 545–625 nm in response to excitation at 542 nm.
Total RNA was isolated from β cells and islets using TRIzol reagent (Thermo Fisher Scientific) in combination with the RNeasy Mini kit (Qiagen, Hilden, Germany) followed by DNase treatment82,83, and cDNA was synthesized via a Thermo-Script RT-PCR System (Thermo Fisher Scientific). After reverse transcription, real-time quantitative PCR was performed on an AbiPRISM 7300 fast real-time cycler using the power SYBR Green real-time PCR master mix kit and quantified by built-in SYBR Green Analysis (Thermo Fisher Scientific)77,84. Samples were measured in triplicates and results were confirmed by at least three independent experiments. The relative amount of specific mRNA was normalized to Glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The sequences of oligonucleotide primers (Merck KGaA, Darmstadt Germany) for gene analysis are listed in Supplementary Table 2.
All results are presented as mean ± s.e.m. Unless otherwise noted, experiments were performed in a blinded fashion at least three times. Statistical analysis was performed via Student’s t test (for 2 groups) unless otherwise indicated, using Prism 7 software (GraphPad, San Diego, CA, USA). A value of P < 0.05 was considered statistically significant.
All data generated or analyzed during this study are included in the present article.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
G.S. is supported by the NIH (K99DK107895) and ES-DRC P30DK20541. We thank Drs. Yaron Tomer, Richard N. Kitsis, Thomas V. McDonald, and Jeffrey E. Pessin (Einstein College of Medicine) for insightful discussion and precious assistance.