Cell Metabolism Article

Type 2 and Congenital Hyperinsulinism Cause DNA Double-Strand Breaks and p53 Activity in b Cells

Sharona Tornovsky-Babeay,1,11 Daniela Dadon,2,11 Oren Ziv,2,11 Elhanan Tzipilevich,2 Tehila Kadosh,2 Rachel Schyr-Ben Haroush,1,2 Ayat Hija,2 Miri Stolovich-Rain,2 Judith Furth-Lavi,1 Zvi Granot,2 Shay Porat,2,3 Louis H. Philipson,4 Kevan C. Herold,5 Tricia R. Bhatti,6 Charles Stanley,6 Frances M. Ashcroft,7 Peter In’t Veld,8 Ann Saada,9 Mark A. Magnuson,10 Benjamin Glaser,1,* and Yuval Dor2,* 1Endocrinology and Metabolism Service, Department of Internal Medicine, Hadassah-Hebrew University Medical Center, Jerusalem 91120, Israel 2Department of Developmental Biology and Cancer Research, The Institute for Medical Research Israel-Canada, The Hebrew University-Hadassah Medical School, Jerusalem 91120, Israel 3Department of Obstetrics and Gynecology, Hadassah-Hebrew University Medical Center, Mount Scopus, Jerusalem 91240, Israel 4Department of Medicine, The University of Chicago, Chicago, IL 60637, USA 5Departments of Immunobiology and Internal Medicine, Yale University School of Medicine, New Haven, CT 06511, USA 6Division of and Diabetes, The Children’s Hospital of Philadelphia, Philadelphia, PA 19104, USA 7Department of Physiology, Anatomy & Genetics, Oxford University, Oxford OX1 3QX, UK 8Diabetes Research Center, Brussels Free University, Laarbeeklaan 103, B1090 Brussels, Belgium 9Monique and Jacques Roboh Department of Genetic Research and the Department of Genetics and Metabolic Diseases, Hadassah-Hebrew University Medical Center, Jerusalem 91120, Israel 10Center for Stem Cell Biology and Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, TN 37232, USA 11These authors contributed equally to this work *Correspondence: [email protected] (B.G.), [email protected] (Y.D.) http://dx.doi.org/10.1016/j.cmet.2013.11.007

SUMMARY INTRODUCTION b cell failure in (T2D) is associated Clinical and experimental data suggest that the pathogenesis of with , but the mechanisms are type 2 diabetes (T2D) involves an initial phase of adaptation to not fully understood. Congenital hyperinsulinism resistance in which pancreatic b cells increase insulin caused by glucokinase mutations (GCK-CHI) is secretion and mass. This is followed by a decompensation associated with b cell replication and apoptosis. phase, where homeostasis is disrupted due to b cell Here, we show that genetic activation of b cell dysfunction and potentially b cell death, resulting in hypergly- cemia and relative hypoinsulinemia (Butler et al., 2003; Weir glucokinase, initially triggering replication, causes et al., 2001). Glucose is thought to play a key role in both phases, apoptosis associated with DNA double-strand initially boosting insulin secretion and b cell replication and later breaks and activation of the tumor suppressor fueling a vicious cycle of glucotoxicity that precipitates b cell p53. ATP-sensitive potassium channels (KATP failure and hyperglycemia (Bensellam et al., 2012; Chang-Chen channels) and calcineurin mediate this toxic et al., 2008; Weir et al., 2001). However, the mechanisms effect. Toxicity of long-term glucokinase over- accounting for this dual effect are not fully understood. Oxidative activity was confirmed by finding late-onset dia- stress, associated with increased metabolic activity, has been betes in older members of a GCK-CHI family. proposed as the driving force behind glucotoxicity (Guo et al., Glucagon-like peptide-1 (GLP-1) mimetic treatment 2013; Poitout and Robertson, 2008); endoplasmic reticulum or p53 deletion rescues b cells from glucokinase- (ER) stress and failure of autophagy have also been implicated induced death, but only GLP-1 analog rescues (Karunakaran et al., 2012; Las and Shirihai, 2010). Congenital hyperinsulinism (CHI) is a rare monogenic disease b cell function. DNA damage and p53 activity in b with clinical manifestations opposite to diabetes (hyperinsuline- T2D suggest shared mechanisms of cell failure mic ), resulting from mutations in key genes in hyperglycemia and CHI. Our results reveal controlling insulin secretion. Most commonly, it results from membrane depolarization via KATP channels, calci- loss-of-function mutations in the Kir6.2 and Sur1 subunits of neurin signaling, DNA breaks, and p53 as determi- the ATP-sensitive potassium channel (KATP channel), encoded nants of b cell glucotoxicity and suggest pharmaco- by KCNJ11 and ABCC8, respectively. These cause permanent logical approaches to enhance b cell survival in b cell membrane depolarization and, consequently, hypersecre- diabetes. tion of insulin (James et al., 2009). Gain-of-function mutations in

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 109 Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

glucokinase (GCK-CHI), the first step in glycolysis, also cause than those in control mice (Figure 1D, middle). However, the excessive insulin secretion, leading to life-threatening hypogly- insulin/glucose ratio, a measure of b cell responsiveness to cemia (Cuesta-Mun˜ oz et al., 2004; Glaser et al., 1998; Kassem ambient glucose levels, was markedly elevated (Figure 1D, right). et al., 2010). We and others have recently shown that islets of Consistent with observations in human GCK-CHI (Kassem GCK-CHI patients are larger than normal (Cuesta-Mun˜ oz et al., et al., 2010) and mice injected with a small-molecule glucokinase 2004; Kassem et al., 2010). In contrast, CHI patients with inacti- activator (Porat et al., 2011), the fraction of b cells engaged in the vating mutations in KCNJ11 or ABCC8 either have small, poorly cell cycle doubled 2 days after transgene activation, as mea- formed islets scattered throughout the (diffuse CHI), or sured by immunostaining and qRT-PCR for cell cycle markers a well-defined, monoclonal lesion characterized as adenoma- (Figure 1E). Thus, this animal model provides a tool to study tous hyperplasia (focal CHI) (Rahier et al., 2011). In all cases, the effect of controlled, increased b cell glycolysis without the affected b cells show increased proliferation (Kassem et al., confounding effects of ambient hyperglycemia. 2000, 2010). These observations, combined with genetic and pharmacologic manipulations of GCK and KATP channels in Prolonged Activation of mutGCK Transgene Causes mice, have highlighted the central role of glycolysis and mem- b Cell Dysfunction and Death brane depolarization in homeostatic b cell replication and its As expected, activation of the transgene resulted in a rapid and potential utility for regenerative therapy in diabetes (Porat significant decrease in random glucose levels (Figure 1F, et al., 2011). However, the islets of CHI patients also contain inset). Surprisingly, however, this decrease was short lived, increased numbers of apoptotic b cells, showing that glucose reaching a nadir at 4 days before returning to normal 1 week after metabolism and membrane depolarization can simultaneously transgene activation (Figure 1F). To determine the reason for exert both mitogenic and toxic effects on human b cells (Kassem this, we quantified the fraction of b cells that expressed the trans- et al., 2000, 2010). gene strongly, using EGFP as a surrogate marker. At 3 days after To better understand how glycolysis controls b cell fate in vivo, tamoxifen injection, 25% of b cells stained for EGFP, reflecting we generated a transgenic mouse model of GCK-CHI. Here, we mosaic transgene expression; 22 days after tamoxifen injection, describe the insights gained from this model on the mechanisms only 5% stained positive (Figure 2A). Since Cre-mediated trans- of b cell glucotoxicity and corroborate these findings in a mouse gene activation is irreversible, disappearance of EGFP+ cells model of diabetes and in tissues from human T2D and CHI suggested that they were gradually eliminated. patients. To examine this more directly, we stained histological sections for apoptosis. As shown in Figure 2B, islets of transgenic mice RESULTS had abundant TUNEL+ b cells compared with minimal staining in control littermates. Furthermore, by day 4 after induction, Acute Induction of GCK Causes Fasting Hypoglycemia the fraction of replicating EGFP+ b cells (measured by Ki67 and Hyperinsulinemia staining) was only 0.34%, compared with 1.29% of b cells in A GCK transgene containing a strong activating mutation found wild-type mice and 1.83% of EGFPÀ b cells in Pdx1-GCK mice in a GCK-CHI patient (Y214C) (Cuesta-Mun˜ oz et al., 2004) was (Figure 2C). These findings suggest that after a few days of activ- inserted into the Rosa26 locus in embryonic stem cells (ESCs), ity, transgene-expressing b cells stop proliferating and are later downstream of a loxP-flanked transcriptional stop cassette (Sri- eliminated. nivas et al., 2001), and followed by an internal ribosome entry site (IRES)-EGFP reporter (Murtaugh et al., 2003). Mice generated b Cell Apoptosis and Late Development of Diabetes in from these cells (mutGCK) were crossed either with Pdx1-CreER Human GCK-CHI mice (Gu et al., 2002) to allow for acute, tamoxifen-induced The findings above show that expression of active GCK in mouse activation of the transgene in adult b cells (Pdx1-GCK) or with b cells is both mitogenic and toxic. A similar phenomenon insulin-Cre mice (Gannon et al., 2000) to force expression of of b cell mitogenicity accompanied by apoptosis, and eventually mutant GCK in b cells as they form in the embryo (ins-GCK) b cell failure and diabetes, was seen in human KATP-CHI (Kassem (Figure 1A). et al., 2000). However, in the case of human GCK-CHI, b cell Islets isolated 3 days after injection of adult Pdx1-GCK mice apoptosis was documented in only a single patient so far (Kas- with tamoxifen showed a 50% increase in the level of GCK sem et al., 2010), and no data are available on the long-term mRNA, indicating expression of the transgene (Figure 1B). outcome of patients. To determine if excessive b cell apoptosis Expression of the mutated GCK transgene is expected to is a general feature of GCK-CHI, we stained archived material enhance glucose utilization and the generation of ATP and, from three previously reported patients carrying different muta- consequently, to boost insulin secretion and reduce blood tions in GCK (pA454 dup and W99L) who underwent pancreatec- glucose. Because of the kinetic characteristics of native and tomy due to uncontrolled hypoglycemia at ages 2–7 years mutant GCK, the largest effect is expected at low glucose levels (Sayed et al., 2009; Snider et al., 2013). Strikingly, all three cases in which the enzymatic activity of native GCK is negligible. showed b cell apoptosis (TUNEL+ insulin+ cells) in 0.59%–1.16% Accordingly, a near doubling of oxygen consumption was of b cells, whereas in 4 controls aged 10 months to 7.8 years, observed in transgenic islets incubated in 2 mM glucose (Fig- TUNEL+ b cells were rarely seen (Figure S1 available online). ure 1C). Activation of the transgene resulted in moderate fasting Apoptosis was more prominent in the two patients with the hypoglycemia 3 days after gene induction (Figure 1D, left). As ex- more severe mutation (pA454 dup). pected from this model of insulin dysregulation, absolute fasting If b cell apoptosis increases, as shown by Kassem et al. (2010) plasma insulin levels were only slightly (but significantly) higher and above, one would expect GCK-CHI patients to enter clinical

110 Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

A Figure 1. A Mouse Model for GCK-CHI (A) Schematic of the transgenic construct. Cre activation causes permanent expression of GCK-Y214C and EGFP tag driven by an internal B C ribosome entry site (IRES). Cell and temporal specificity is determined by the Cre-expressing transgene used. (B) Expression level of GCK in islets isolated from Pdx1-GCK mice 3 days after tamoxifen- induced expression of the transgene by quan- titative RT-PCR. GCK levels (including both endogenous and transgenic) were normalized to b-actin. n = 5 mice/group. Control mice were single transgenic littermates carrying either the D GCK or Cre transgene. (C) Increased oxygen consumption in islets isolated from Pdx1-GCK mice 2 days after tamoxifen-induced transgene expression and incubated at 2 mM glucose. n = 10 control and 9 mutGCK mice. OCR, oxygen consumption rate. (D) Fasting hypoglycemia (left), fasting hyper- insulinemia (center), and increased insulin/ glucose ratio (right; ratio 3 100) in 5-week-old E Pdx1-GCK mice 3 days after tamoxifen injection. n = 12 control and 11 Pdx1-GCK mice. (E) The fraction of replicating b cells doubled in Pdx1-GCK mice 2 days after transgene induc- tion. Left: quantification of Ki67+ insulin+ cells in paraffin sections. n = 6–8 mice/group, >2,000 b cells counted/mouse. Right: qRT-PCR for cell cycle markers on RNA extracted from islets 2 days after tamoxifen injection. n = 4 mice/ group. (F) Hypoglycemia after induction of Pdx1-GCK F transgene is transient. Tamoxifen was injected at day 0. n = 21 Pdx1-GCK and 12 control mice. In each time point, at least 5 mice/group were measured. Data are represented as mean ± SE.

remission (cessation of hypoglycemic episodes) and perhaps Pdx1-GCK mice, even though the time to b cell failure is dramat- even develop diabetes over time. We contacted one previously ically shorter in the mouse model (days versus years). reported GCK-CHI family (Glaser et al., 1998) and found that of five mutation carriers, one was diagnosed with insulin-requiring Molecular Characterization of b Cell Glucotoxicity diabetes at age 48, as reported in the original publication; the To understand the basis of the observed toxicity of active GCK, proband had diabetes diagnosed at age 39 and is currently we examined the expression of multiple markers of b cell identity treated with oral medications; his sister died of breast cancer and stress after tamoxifen-induced expression of mutant GCK in at age 49 without diabetes; and the proband’s son has had adult Pdx1-GCK mice. Staining for the key b cell transcription -treated diabetes since age 28 (BMI 26.6); whereas factors MafA and Nkx6.1, but not Pdx1, was dramatically the proband’s daughter’s hypoglycemia remitted at age 14 reduced, as shown to occur in high glucose levels in vitro (Kaneto (currently euglycemic; BMI 30.7). Thus, three of the four living et al., 2008; Weir et al., 2001) and in mouse and human T2D (Guo mutation carriers developed diabetes, and one is in remission, et al., 2013)(Figure S2). In addition, we stained for the ER, oxida- suggesting reduced b cell function. The deterioration of b cell tive, or genotoxic stress marker Chop (also known as Gadd153 function over time may not occur in all families or all mutations, or Ddit3). Approximately 0.6% of cells in transgenic islets stained and a systematic review is needed to fully define the long-term for Chop compared with none in wild-type littermates (Figure 3A, effect of activating GCK mutations in man. left and middle panels). Consistent with these findings, qRT-PCR In summary, evidence from human GCK-CHI indicates exten- analysis revealed increased expression of Chop mRNA (Fig- sive b cell apoptosis giving rise with time to b cell failure and ure 3A, right). Atf3, another ER and oxidative stress marker diabetes. This conclusion is consistent with the phenotype of (Eizirik et al., 2008), was also markedly elevated in Pdx1-GCK

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 111 Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

Figure 2. Apoptosis and Reduced Prolifera- A Ins Pdx1 GFP b GFP tion of Cells following Expression of Pdx1 Active GCK *** (A) Gradual loss of cells expressing high levels of 30 GFP in Pdx1-GCK mice. Upper panel, 3 days after 3 days 25 tamoxifen; n = 5 mice. Lower panel, 22 days; n = 4 20 mice. Scale bars: 20 mm (here and in all figures + b 15 except Figure 5B). Quantification of GFP cells at beta cells

+ days 3 and 22 is on the right. 10 (B) TUNEL staining showing apoptosis of Pdx1- 5 GCK b cells and quantification. High magnification %GFP

22 days 0 in the lower panels demonstrates insulin staining 3 days 22 days of TUNEL+ cells. Pdx1-GCK mice were injected with tamoxifen 4 days prior to sacrifice. n = 4–5 mice/group. (C) Reduced replication of GFP+ b cells compared B with control and GFPÀ cells 4 days after tamoxifen Control injection to Pdx1-GCK mice. n = 12 Pdx1-GCK and 8 control mice. Data are represented as mean ± SE. See also 1.6 *** Figure S1. 1.4 Ins TUNEL 1.2 DNA TUNEL DNA 1 beta cells Pdx1-GCK + 0.8 0.6 the mutation used in our transgenic 0.4 mice) (Kassem et al., 2010). Abundant 0.2 gH2AX-stained b cells were seen in islets %TUNEL 0 of this patient, compared with no gH2AX WildControl type Pdx1-GCK staining in age-matched controls. This confirms relevance in man and indicates that DNA damage results from increased C GCK activity and is not a consequence of *** 2.5 a specific GCK mutation (Figure 3D). 2 *** DNA damage typically results in activa- 1.5 Insulin tion of the tumor suppressor p53, which 1 beta cells + 0.5 may induce cell-cycle arrest and a DNA 0 damage response or trigger programmed wild type GFP mutGCK- cells GFPmutGCK+ cells %Ki67 cell death (Halazonetis et al., 2008). Strik- Control GFP- cells GFP+ cells TUNEL DNA Pdx1-GCK ingly, 2.3% of the nuclei in Pdx1-GCK islets stained positive for p53, compared with 0.4% in controls (Figure 3E). islets (Figure 3A, right), whereas the levels of spliced Xbp1, a These findings suggest that excessive glucose metabolism in more specific marker of ER stress, were unchanged (Figure 3A, b cells leads, perhaps via oxidative stress and reactive oxygen right panel). This suggests that the cause of elevated Chop and species (ROS), to double-strand breaks in the DNA, activation Atf3 is oxidative or genotoxic stress, consistent with the idea that of p53, and apoptosis. We next studied the signaling pathway oxidative stress is a major determinant of b cell failure in T2D leading from glycolysis to DNA damage. (Guo et al., 2013). Oxidative stress was previously shown to cause double- b Cell Membrane Depolarization Is Necessary and strand DNA breaks (Cadet et al., 2012). Strikingly, 13.8% of Sufficient to Trigger DNA Damage b cell nuclei in transgenic mice contained foci stained for Increased glucose metabolism in b cells results in increased 53BP1, a protein recruited to sites of double-strand breaks in mitochondrial ATP production. In the insulin secretion pathway,

DNA (Schultz et al., 2000), compared with only 1.2% in controls this causes the closure of KATP channels, membrane depolariza- (Figure 3B). Consistently, 1.75% of nuclei in transgenic islets tion, elevated intracellular calcium concentrations, and ulti- stained for gH2AX, a histone variant marking chromatin areas mately insulin exocytosis. To study how expression of mutant flanking double-strand DNA breaks (Rogakou et al., 1998), GCK results in b cell DNA damage, we asked if increased glycol- compared with only 0.2% in controls (Figure 3C). Increased ysis acted through membrane depolarization as in insulin secre- gH2AX was also seen in islets that were isolated from Pdx1- tion. We utilized transgenic mice expressing a Cre-inducible GCK mice after tamoxifen injection and cultured for 3 days, sup- constitutively active mutant of Kir6.2, a subunit of the b cell porting the cell-autonomous nature of the effect (Figure S3). To KATP channel. Expression of the Kir6.2-V59M mutation in b cells determine whether the presence of DNA damage is specific to prevents membrane depolarization and, consequently, blocks this mouse model, we examined pancreatic sections from a glucose-stimulated insulin secretion and causes diabetes (Gir- patient with an activating GCK mutation (V91L, different than ard et al., 2009). Acute expression of the Kir6.2 transgene was

112 Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

Figure 3. Molecular Markers of b Cell Glu- wildControl type mutGCKPdx1-GCK Control Pdx1-GCK cotoxicity in Mice and Humans A 4 ** ** 3.5 (A) Left: immunostaining for the stress marker 3 Chop in islets of control and Pdx1-GCK mice 2.5 4 days after tamoxifen injection. Right: qRT-PCR 2 1.5 analysis of Chop and Atf3 (normalized to b-actin) 1 and spliced Xbp1 (normalized to total Xbp1) in Fold change 0.5 RNA from Pdx1-GCK islets 3 days after tamoxifen Ins CHOP 0 injection. n = 5 mice/group. Arrows point to Chop+ CHOP ATF3 s-xbp1/t-xbp1 b cells. (B) Staining for the DNA damage marker 53BP1 in Control Pdx1- B 20 *** b cells of Pdx1-GCK mice 4 days after tamoxifen GCK injection. Right panel shows the frequency of 15 b cells containing nuclear foci of 53BP1. n = 6 mice/group, 1,800–2,700 cells counted/mouse. 10

beta cells +

+ Arrows point to 53BP1 b cells. 5 (C) Staining and quantification for the DNA dam- age marker gH2AX in b cells of control and Pdx1-

%53BP1 0 GCK mice 4 days after tamoxifen injection. n = 4 Ins 53BP1 Pdx1 wildControl type Pdx1-GCK mutGCK Pdx1-GCK and 3 control mice. There were 2,000– 3,500 cells counted/mouse. Arrows point to gH2AX+ b cells. C Control Pdx1-GCK D 2 (D) Staining for gH2AX in pancreatic sections from *** hGCK-CHI a 36-month-old human GCK-CHI patient (V91L) Ins (Kassem et al., 2010). Arrows point to gH2AX+ 1.5 H2AX b cells. Pdx1 (E) Staining and quantification for p53 in islets of

beta cells 1 + control and Pdx1-GCK mice 4 days after tamox- ifen injection. n = 3 control and 4 Pdx1-GCK mice. 0.5 H2AX Arrows point to p53+ islet cells.

% Data are represented as mean ± SE. See also 0 Figures S2 and S3. Ins H2AX Pdx1 Control Pdx1-GCK

E Control Pdx1-GCK 3.5 * 3 2.5 controls (Figure 4B). To confirm that this 2 difference was due to the mutations and

islet cells 1.5 + not to other differences between patients 1 and controls, we examined gH2AX and %p53 0.5 53BP1 in a patient with the focal form of 0 CHI. In this disease variant, the patient p53 p53 Control Pdx1-GCK inherits a single, recessive, inactivating

KATP channel gene mutation, and a sec- achieved by tamoxifen injection of adult Pdx1-CreER;Kir6.2- ond somatic mutation results in loss of the normal allele and

V59M mice (Pdx1-Kir6.2). As expected, these mice rapidly clonal expansion of b cells carrying only an inactivating KATP developed severe hyperglycemia (Figure 4A, left). Compound channel mutation. This causes the formation of a discrete focal transgenic mice coexpressing both Kir6.2-V59M and mutant lesion containing defective, hyperfunctioning b cells, whereas GCK in b cells (Pdx1-GCK;Kir6.2) were also hyperglycemic b cells outside the lesion function normally (Glaser et al., 1999). (like Pdx1-Kir6.2 mice), reflecting the fact that membrane depo- This allows comparison between defective and normal b cells larization is necessary for insulin secretion downstream of GCK that were exposed to precisely the same milieu prior to resection. (Figure 4A, left). b cells within the focal lesion stained positive for gH2AX and had Strikingly, expression of Kir6.2-V59M reduced gH2AX staining 53BP1 nuclear foci, whereas cells outside the lesion did not (Fig- in b cells of Pdx1-GCK mice 4 days after tamoxifen injection ure 4B). We have previously shown increased b cell apoptosis in (Figure 4A, middle) and attenuated b cell apoptosis as judged patients with both diffuse and focal CHI caused by the same by TUNEL staining (Figure 4A, right). These results suggest gene mutations (Kassem et al., 2000). These findings show that membrane depolarization is necessary for glycolysis- that b cell membrane depolarization is both necessary and suffi- induced DNA double-strand breaks and apoptosis in b cells. cient to cause the DNA damage response seen following GCK We then asked if membrane depolarization, in the absence of activation. increased glycolysis, is sufficient to cause DNA double-strand breaks in b cells. We stained sections of pancreata resected Calcineurin Mediates b Cell Apoptosis Downstream of from patients with KATP-CHI, caused by mutations in either Active GCK ABCC8 or KCNJ11. Both gH2AX and 53BP1 staining were In the insulin secretion pathway, membrane depolarization is observed in b cells of these CHI patients, but not in age-matched immediately followed by the opening of voltage-dependent

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 113 Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

A Figure 4. Chronic Membrane Depolarization and Calcineurin Mediate b Cell Glucotoxicity (A) Analysis of Pdx1-GCK mice expressing a Kir6.2- V59M mutation that prevents membrane depolari- zation. Left: hyperglycemia develops in Pdx1- Kir6.2 mice regardless of the presence or absence of mutated GCK. Center: the Kir6.2-V59M mutation reduces DNA double-strand breaks in Pdx1-GCK mice 4 days after tamoxifen injection. Right: the Kir6.2-V59M mutation reduces b cell apoptosis in Pdx1-GCK mice. n = 6–10 mice/group. (B) Immunofluorescence detection of the DNA damage response markers gH2AX and 53BP1 in paraffin sections of human pancreas. Control: a healthy individual (similar results were obtained at 1 or 3 months old). Kir6.2 diffuse: patient was B 3 months old and homozygous for KCNJ11 p.(Tyr12*). Sur1 diffuse: patient was 6 months old and homozygous for ABCC8 p.(Arg836*; similar results were found for a patient aged 1 month that was homozygous for the ABCC8 c.950 delC mutation). Sur1 ‘‘focal’’ and ‘‘outside’’ refer to areas within the same pancreas, inside a hyperplastic focal lesion and outside it, in the normal adja- cent pancreas (paternal heterozygous ABCC8 p.[Arg836*]; age 70 months). All patients and con- trols were previously reported (Kassem et al., 2000). (C) Treatment of Pdx1-GCK mice with the calci- neurin inhibitor tacrolimus (FK-506, FK) reduces b cell apoptosis 4 days after tamoxifen injection. n = 4–7 mice/group. Mice were injected with either 0.3 mg/kg or 1 mg/kg FK-506/day from the day of C tamoxifen injection to sacrifice. For each mouse, 2,300–8,000 b cells were counted. Data are represented as mean ± SE.

of p53 in glucokinase-induced b cell death. To accomplish this, we generated mice in which activation of mutGCK in b cells was coupled to Cre-mediated dele- tion of p53 (Marino et al., 2000). calcium channels (VDCCs). To ask if VDCCs mediate the toxic To maximize the efficiency of recombination, we crossed insu- effects of active GCK, we treated Pdx1-GCK mice with the lin-Cre mice with mutGCK mice. In the offspring (ins-GCK mice), calcium channel blocker nifedipine. As expected, this led to Cre-mediated recombination occurs in b cells during embryonic transient hyperglycemia due to blockade of insulin secretion. development, immediately after the onset of insulin expression. However, histological examination of pancreata 4 days after At day 16.5 of gestation, ins-mutGCK embryos had reduced activation of the GCK transgene did not reveal a significant b cell area, and insulin+ cells stained frequently for gH2AX (Fig- decrease of b cell apoptosis or DNA damage (not shown). To ure S4). At birth, transgenic mice were hyperglycemic (Figure 5A, further examine how calcium signaling affects glucotoxicity, left), mildly hypoinsulinemic (Figure 5A, middle), and their insulin/ we treated Pdx1-GCK mice with tacrolimus (FK-506), a selec- glucose ratio was lower than that of controls (Figure 5A, right). tive inhibitor of calcineurin, a phosphatase activated by b cell mass was reduced by >75% compared with controls (Fig- calcium. As shown in Figure 4C, tacrolimus significantly reduced ure 5B), and 80% of the mice died during the first week of life b cell apoptosis, suggesting that calcium signaling via calci- (see below). These observations are consistent with those neurin activity is required for the toxic effects of glucose. The described above for Pdx1-GCK mice. The more severe pheno- source of the calcium (intra- or extracellular) and the identity of type probably results from highly efficient recombination driven the responsible calcium channel remain to be discovered. by the insulin-Cre transgene and perhaps from an increased sensitivity of newborn b cells to glucotoxic injury. Glucose-Induced b Cell Death, but Not Dysfunction, Is Loss of p53 in b cells (achieved by incorporation of loxP- Mediated through p53 flanked p53 alleles, p53lox/lox) preserved islet architecture and Because oxidative stress can induce apoptosis through a p53- b cells in newborn ins-GCK mice (Figure 5C), suggesting preven- dependent mechanism, and since b cells expressing the trans- tion of GCK-induced b cell death. However, ins-GCK;p53lox/lox gene express p53, we next examined the functional significance mice were only slightly less hyperglycemic than age-matched

114 Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

ins-GCK mice (which have normal p53) (Figure 5D). This finding Finally, we stained pancreatic sections from human cadaver is expected because p53 deficiency can prevent cell death, but organ donors with T2D. Extensive 53BP1 foci were observed in does not interfere with upstream cell dysfunction. To determine if b cells from all four T2D donors examined, regardless of age or p53 deficiency cell autonomously rescued transgene-express- disease duration (7.3% versus 1.1% of b cells in T2D and ing b cells, we stained for EGFP. In ins-GCK mice that survived controls, respectively) (Figure 7D and Table S1). Staining for to 3 weeks of age, no EGFP+ cells were observed, suggesting gH2AX and p53 was not seen (not shown). These findings sug- that cells strongly expressing the GCK mutation died and that gest that double-strand breaks in DNA and downstream activa- transgene-negative b cells took over. In contrast, at this age, tion of p53 may be a common feature of all forms of metabolic many islets of ins-GCK;p53lox/lox mice stained strongly for b cell stress, whether caused by hyperactivity driven by specific EGFP, proving a cell-autonomous rescue of transgene-positive gene mutations, as in CHI, or by hyperglycemia, as in diabetes. b cells (Figure 5E). Finally, p53 deficiency in b cells increased the chance of ins-GCK mice surviving postnatally: only 20% of DISCUSSION ins-GCK mice survived to 3 weeks of age, compared to 60% of ins-GCK;p53lox/lox mice (Figure 5F). Double-Strand Breaks and p53 Activation in Glucotoxic These results suggest that p53 deficiency can prevent gluco- b Cell Failure toxic b cell death, but it does not interfere with upstream meta- Our results identify a molecular pathway, conserved from mouse bolic processes that cause b cell dysfunction. In newborn to man, that mediates the toxic effects of glucose on pancreatic mice, the residual activity of surviving b cells appears to be suf- b cells. We showed that increased glucose metabolism in b cells ficient to keep most animals alive. causes double-strand breaks in DNA, likely accounting for acti- vation of the p53 tumor suppressor and triggering b cell death. GLP-1 Treatment Prevents Both b Cell Death and This series of events is cell autonomous and triggered not by Dysfunction hyperglycemia per se, but by the rate of intracellular glucose We next attempted a pharmacological rescue of b cell function. metabolism. We also demonstrate that membrane depolariza-

GLP-1 analogs alleviate both ER stress (Yusta et al., 2006) and tion via closure of KATP channels mediates the glucose-induced oxidative stress (Shimoda et al., 2011)inb cells. Liraglutide, a DNA damage response and apoptosis, with involvement of GLP-1 receptor agonist and a T2D drug, was administered to calcineurin signaling downstream. Experiments using inhibitors adult tamoxifen-injected Pdx1-GCK mice. At 4 days after trans- of VDCCs failed to rescue glucotoxicity in mutGCK islets, making gene induction, liraglutide-treated animals had no Chop+ GFP+ it difficult to determine a role of these channels. b cells (Figure 6A). The fraction of gH2AX+ b cells was reduced The identification of double-strand DNA breaks and p53 acti- (Figure 6B), as was b cell apoptosis (Figure 6C). In addition, lira- vation in b cells in both CHI and T2D suggests that a similar glutide treatment neutralized the negative effect of mutant GCK mechanism operates in these seemingly opposite diseases, expression on b cell replication, observed 4 days after tamoxifen regardless of whether excessive b cell activity is driven by muta- injection (Figure 6D). Consistent with the reduction in b cell tions in GCK or the KATP channel (as in CHI) or by chronic hyper- apoptosis, mice treated with liraglutide for 9 days after tamoxifen glycemia (as in T2D). We propose that DNA double-strand had more GFP+ cells compared with untreated littermates (Fig- breaks and p53 are central players in glucose-induced b cell fail- ure 6E). In contrast to the effect of p53 deficiency, liraglutide ure in both hyperinsulinism and T2D. Other evidence supporting treatment rescued b cell function. Pdx1-mutGCK mice treated a link between DNA damage and b cell failure includes the with liraglutide showed sustained hypoglycemia (Figure 6F), as demonstration of diabetes in mice lacking the Werner syndrome predicted. As expected from the known mechanism of action DNA helicase (Moore et al., 2008), an increased incidence of dia- of the drug, liraglutide did not cause hypoglycemia in control betes in individuals who received irradiation to the pancreas (de animals. Vathaire et al., 2012; Meacham et al., 2009), and identification of oxidized DNA in islets of T2D patients (Sakuraba et al., 2002). DNA Damage and p53 Activity in Type 2 Diabetes Analysis of islets from T2D patients previously identified an To determine if the mechanisms responsible for glycemic b cell increased expression of proteins in the p53 pathway (Nyblom stress elucidated here are also operative in the context of chronic et al., 2009) and increased mRNA expression of the p53 target hyperglycemia, we studied pancreatic specimens from diabetic p21 (Marselli et al., 2010). Like most cell types, b cells are sensi- mice and humans. tive to activation of p53, as demonstrated by the development of db/db mice lack the leptin receptor and are hyperphagic, b cell failure in transgenic mice overexpressing a superactive p53 obese, insulin resistant, and hyperglycemic. In 12-week-old mutant (Hinault et al., 2011). We note, however, that the roles of hyperglycemic db/db mice, 0.88% of b cells were positive for DNA double-strand breaks and p53 in the pathogenesis of T2D gH2AX, compared with 0.07% in controls (Figure 7A). 53BP1 remain unproven. It is also possible that the driver for DNA dam- nuclear foci were seen in 9.2% of db/db b cells, compared age in T2D is not excessive glycolysis as in CHI, given a report with 1.2% in controls (Figure 7B). Islets in db/db pancreata that oxygen consumption is reduced rather than enhanced in also had cells (2.3%) that stained positively for p53, compared T2D islets in vitro (Doliba et al., 2012). with 0.4% in controls (Figure 7C), demonstrating a striking simi- A potential caveat in this study is that both Pdx1-CreER and larity to b cells of Pdx1-mutGCK and ins-mutGCK mice. Further- insulin-Cre mice drive recombination in some non-b cells, more, expression of the p53 target gene p21 (CIP1/WAF1) was including the hypothalamus (in both lines), and in acinar cells, increased in b cells of db/db mice compared with controls, indi- duodenal cells, and delta cells (in Pdx1-CreER) (Wicksteed cating biological activity of p53 (Figure S5). et al., 2010). To rule out confounding effects of extra-islet

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 115 Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

A 0.8 NS 1 * 500 *** 0.8 400 0.6 0.6 300 0.4 200 0.4 0.2 100 Insulin ng/ml 0.2 Insulin/Glucose Glucose (mg/dL) 0 0 0 Control Ins-GCK Control Ins-GCK Control Ins-GCK

B Control Ins-GCK 2.5 *** 2 1.5 1 0.5

% area 0 Control Ins-GCK Ins Ins

C Control Ins-GCK Ins-GCK;p53lox/lox D

P3 Ins-GCK Ins Glu+Soma Ins Glu+Soma Ins Glu+Soma Control Ins-GCK p53lox/lox E Ins-GCK 3 weeks Ins-GCK;p53lox/lox 3 weeks

Ins GFP GFP Ins GFP GFP

F p53KO rescues Ins-GCK mice 60 *** NS

40

20

% Ins-GCK 0 p53 wt p53lox/lox Expected Observed

Figure 5. p53 Mediates GCK-Induced b Cell Toxicity (A) Activation of mutGCK in b cells during embryonic development (ins-GCK) results in hyperglycemia (left), mild hypoinsulinemia (center), and decreased insulin/ glucose ratio (right; ratio 3 100) in postnatal day 3 mice. n = 24 control and 10 ins-GCK mice. (B) Reduced b cell area in postnatal day 3 ins-GCK mice and representative images. Images were taken at 403 magnification and stitched together to generate high-resolution pictures of whole pancreatic sections (scale bar: 200 um). n = 3 mice/group. Insets show a higher magnification to better visualize insulin-stained islets (scale bar: 100 um). (C) b cell-specific p53 deletion preserves b cells and islet structure in postnatal day 3 ins-GCK mice. Green, insulin; red, glucagon + somatostatin. (D) Blood glucose levels in newborn mice expressing the GCK mutation in b cells, with or without deletion of p53 (n > 3 mice/group). (E) Expression of GFP in islets of 3-week-old ins-GCK mice, with (left) or without (right) codeletion of p53. No GFP+ cells were seen in surviving ins-GCK mice. The presence of GFP+ b cells in ins-GCK mice deficient for p53 demonstrates the cell-autonomous rescue of transgene-positive cells. (legend continued on next page) 116 Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

Figure 6. Pharmacologic Rescue of GCK- A B *** 1 * 2 * Induced b Cell Toxicity

+ Pdx1-GCK mice were treated twice a day with 0.8 1.5 liraglutide, starting 1 day after tamoxifen injection GFP / until sacrifice at day 4 (A–D) or day 9 (E and F). + 0.6 + beta cells

+ 1 (A) Elimination of Chop in GFP b cells (left). n = 3–4 0.4 mice/group. 0.5 H2AX (B) Reduced gH2AX in b cells of Pdx1-GCK mice %CHOP 0.2 treated with liraglutide. n = 3–4 mice/group. % 0 0 (C) Reduced b cell apoptosis in b cells of Pdx1-GCK Pdx1-GCK Pdx1-GCK+ Control Pdx1-GCKmutGCK Pdx1-GCK+ Liraglutide Liraglutide mice treated with liraglutide. n = 4–6 mice/group. (D) Rescued replication of GFP+ b cells in Pdx1-GCK mice 4 days after injection of tamoxifen. Filled bars, CD GFPÀ b cells; hatched bars, GFP+ b cells represent- 2.5 ** ing cells expressing a high level of the GCK trans- 1.6 ** *** 2 gene. n = 6–11 mice/group. 1.4 (E) Improved survival of b cells strongly expressing 1.2 1.5 1 GFP. Liraglutide was administered twice a day to beta cells Beta cells

+ 1

0.8 + mice for 8 days, starting 1 day after tamoxifen 0.6 0.5 injection until sacrifice. n = 7–11 mice/group. 0.4 (F) Sustained hypoglycemia in Pdx1-GCK mice

0.2 %Ki67 0 - + - + treated with liraglutide. n = 7–11 mice/group. Blood

%TUNEL 0 GFP GFP GFP GFP Control Pdx1-GCKmutGCK Pdx1-GCK + Pdx1-GCK + glucose levels of control mice and control mice Liraglutide Control Pdx1-GCK Liraglutide injected with liraglutide were not statistically different. Pdx1-GCK mice injected with liraglutide 45 had significantly lower (p < 0.05) blood glucose levels E * F 150 40 compared with Pdx1-GCK mice at days 2, 8, and 9. 35 Data are represented as mean ± SE. 30 100 25

beta cells 20 + Control 15 50 Control + Liraglutide poglycemia and so are expected to have 10 Pdx1-GCK %GFP Glucose (mg/dl) Pdx1-GCK + Liraglutide normal or reduced glycolysis. Our findings 5 0 0 0 1 2 3 4 5 6 7 8 9 are consistent with evidence that treatment Pdx1-GCK Pdx1-GCK + Liraglutide Days post injection with KATP channel openers (forcing hyper- polarization and causing b cell rest) protects b cells against a number of toxic insults expression of the transgene, we isolated islets from adult Pdx1- (Maedler et al., 2004; Ritzel et al., 2010), although in these GCK mice 18 hr after tamoxifen administration and cultured studies the roles of DNA damage and p53 were not recognized. them in vitro. After 3 days of culture, Pdx1-GCK islets had Our results are also in line with reports of increased apoptosis more gH2AX+ b cells than control islets (Figure S3), supporting and reduced b cell mass in mice lacking Kir6.2 (Miki et al., 2001). the cell-autonomous effect of GCK on b cell DNA damage. The Second, the importance of membrane depolarization sug- finding of DNA damage and b cell apoptosis in embryonic day gests (but does not prove) that calcium may be involved in 16.5 (E16.5) ins-GCK embryos (Figure S4) further indicates that mediating glycolysis-induced DNA damage, p53 activation, the key driver of b cell stress is intracellular glucose metabolism, and apoptosis as previously suggested (Wang et al., 2011). Inter- regardless of circulating glucose levels, which in the embryo are estingly, DNA damage and p53 activation are implicated in controlled by the mother. neuronal excitotoxicity and degeneration (Miller et al., 2000; Neema et al., 2005; Sakhi et al., 1994), and a recent study Signaling from Glucose to DNA Damage, p53 Activation, showed that double-strand breaks form after normal neuronal and b Cell Death activity and are increased in a model of Alzheimer’s disease Our experiments do not completely resolve the signaling path- (Suberbielle et al., 2013). We speculate that b cell glucotoxicity way leading from glucose metabolism to DNA damage and is similar to neuronal excitotoxicity with respect to the crucial b cell death. However, the finding that membrane depolarization role of calcium overload. It will be interesting to examine the rele- via KATP channels is both necessary and sufficient in this vance for diabetes of pharmacologic approaches developed to pathway provides some clues. First, glucotoxicity is not a direct minimize neuronal damage after stroke (Greig et al., 2004). Phar- consequence of excessive glycolysis or cellular energetics. macologic inhibition of calcineurin, a calcium-activated phos-

Indeed, b cells in KATP-CHI patients are exposed to chronic hy- phatase, rescued b cell apoptosis in Pdx1-GCK mice. This is

(F) Ins-GCK mice at 3 weeks of age as a percentage of the total number of mice in the litter. When insulin-Cre+/À (heterozygous) and GCK homozygous mice are crossed, 50% of progeny are expected to contain both transgenes (black bar). On the wild-type p53 background (blue bar), we obtained 39 adult mice from 6 litters and observed only 4 ins-GCK mice (binomial test, p < 0.0001). On the p53-deleted background (red bar), we obtained 11 mice from 7 litters and observed 4 ins-GCK mice (binomial test, p = 0.3). Data are represented as mean ± SE. See also Figure S4.

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 117 Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

Figure 7. b Cell DNA Double-Strand Breaks A B C and Activation of p53 in Type 2 Diabetes (A) Staining for gH2AX and quantification in islets of diabetic db/db mice. Control mice were age- matched wild-type mice. n = 3 db/db and 4 control mice. Arrows point to gH2AX+ b cells. (B) Staining for 53BP1 in islets of diabetic db/db mice and quantification. n = 3 db/db and 6 control mice. Arrows point to b cells with nuclear 53BP1+ foci. Bars represent the percentage of b cells containing at least one 53BP1+ nuclear focus. (C) Staining for p53 and quantification in islets of diabetic db/db mice. n = 5 db/db and 3 control mice. (D) Staining and quantification of 53BP1 nuclear foci in pancreatic sections of human organ donors with or without T2D. n = 4/group. Data are represented as mean ± SE. See also Figure S5 and Table S1.

in CHI. The latter is clearly seen in the Pdx1-GCK mouse model. In the case of

KATP-CHI, patients who do not undergo pancreatectomy eventually enter clinical remission, whereas those who do un- D dergo partial pancreatectomy develop insulin-requiring diabetes, often many years after surgery (Beltrand et al., 2012; Huopio et al., 2003; Leibowitz et al., 1995). In the case of GCK-CHI, we showed here b cell apoptosis as well as long-term remission and progression to diabetes in multiple cases. We note the striking differences in timing of dis- ease progression, which mask the quali- tative similarity: human T2D typically develops over decades, and diabetes from GCK-CHI also occurs only late in life, while Pdx1-GCK mice switch from hypoglycemia to b cell failure in just a reminiscent of the neuroprotective effects of calcineurin inhibi- few days. The latter likely reflects distinct technical aspects of tors (Nito et al., 2011) and further underscores the similarity the transgenic system, e.g., the presence of three GCK alleles between b cell glucotoxicity and neuronal excitotoxicity. Further and expression of mutant GCK from an ectopic promoter not experiments are needed to pinpoint the signaling events leading responding to glucose. from KATP channel closure to DNA damage and, consequently or independently, to b cell apoptosis. Glucose-Induced b Cell Replication versus Death More work is required to identify where the glycolysis-mediated Similarity between Type 2 Diabetes and Congenital mitogenic and toxic pathways diverge and the reason for domi- Hyperinsulinism nance of one outcome over the other. From a therapeutic While T2D and CHI present with opposite glycemia phenotypes, perspective, an outstanding question is whether small-mole- our results suggest that these diseases share some pathogenic cule glucokinase activators (GKA), shown to increase insulin mechanisms. In both, b cells initially respond to increased glycol- secretion and b cell replication (Nakamura et al., 2009; Porat ysis or membrane depolarization by enhanced insulin secretion et al., 2011; Shirakawa et al., 2013), may also be toxic to and replication. This allows the maintenance of euglycemia in b cells. The latter is suggested by the inability of GKA to insulin-resistant prediabetics and causes pathologic hypoglyce- increase b cell mass (Nakamura et al., 2009, 2012) and the mia in CHI patients. At later stages, b cell glucotoxicity becomes transient benefit reported in clinical trials (Kiyosue et al., dominant, at least in part via the molecular pathway discussed 2013; Meininger et al., 2011; Wilding et al., 2013). It will also above. The ensuing b cell failure causes hyperglycemia in T2D, be important to determine the primary stress impacting b cells, and the normalization of glycemia followed by hyperglycemia, as this may have clinical implications (for example, regarding

118 Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

the use of antioxidants if oxidative stress emerges as a culprit) Oxygen Consumption (Guo et al., 2013). Nonetheless, our current data suggest Oxygen consumption rate was measured using a Seahorse XF24 (Seahorse that the mitogenic and toxic effects of glucose on b cells Bioscience). Islets were isolated from control or mutGCK mice 2 days after tamoxifen injection and incubated overnight at 5.5 mM glucose. Pools of can in principle be uncoupled pharmacologically (in our case, Pdx1-GCK and control islets were divided into groups of sixty islets and plated using a GLP-1 angonist), providing a positive outlook for in Seahorse XF24 Islet Capture Microplates. Basal respiration at 2 mM glucose regenerative therapy in diabetes. Combined treatment with was measured. Low glucose levels were selected to optimally observe the

GKA and GLP-1 analogs (Nakamura et al., 2012) may be a first activity of mutant GCK, which has a lower KM to glucose than wild-type step toward productive, nontoxic, glycolysis-induced b cell GCK. Values were normalized to islet area (images of islets taken after oxygen replication. measurements) and analyzed by Seahorse software.

RT-PCR EXPERIMENTAL PROCEDURES Islets were isolated from Pdx1-GCK mice 2 days (for replication markers) or 3 days (for stress markers) after tamoxifen injection. Total RNA was isolated, Mice and cDNA synthesis was performed using random primers (Roche) and To generate mutGCK mice, the Y214C GCK mutation was introduced into rat reverse transcriptase (ImProm-II; Promega). For quantitative real-time PCR, GCK cDNA (a gift from Donald K. Scott) and cloned upstream of IRES-EGFP we used SYBR Green Master Mix in the 7900HT instrument in triplicate and downstream of a loxP-flanked stop cassette in a plasmid containing (Applied Biosystems). The relative amount of mRNA was calculated using Rosa26 homology arms (Murtaugh et al., 2003; Srinivas et al., 2001). After comparative CT method after normalization to b-actin. PCR primers are sequence validation, the plasmid was electroporated into ESCs, and colonies described in Table S3. were screened for homologous recombination. Targeted cells were injected to C57BL/6 blastocysts, and resulting chimeras were bred for germline transgene transmission. Statistical Analyses Blood glucose was measured as described (Nir et al., 2007). Serum insulin In all graphs, *p < 0.05; **p < 0.01; ***p < 0.005; NS, p > 0.05 by Student’s t test, was measured using an ELISA kit (Crystal Chem) after overnight fasting (adult unless otherwise stated (Figure 5F). Data are represented as mean ± SE. mice, Figure 1) or without fasting (postnatal day 3 pups, Figure 5). Liraglutide (Novo Nordisk) was dissolved in saline, and 100 ug/kg was injected subcuta- SUPPLEMENTAL INFORMATION neously twice per day (compared with injection of saline). Tamoxifen (Sigma) was dissolved in corn oil (Sigma) to 20 mg/ml, and 8 mg was injected subcu- Supplemental Information includes five figures and three tables and can be taneously at days 0 and 2. For in vitro experiments (Figure S3), tamoxifen was found with this article online at http://dx.doi.org/10.1016/j.cmet.2013.11.007. injected once. Tacrolimus (FK-506) (Prograf; Astellas) was dissolved in phos- phate-buffered saline (PBS) to 0.1 mg/ml, and 1 mg/kg was injected intraper- itoneally daily from day 1 after tamoxifen administration. Nifedipine (Sigma) ACKNOWLEDGMENTS was freshly dissolved in 5% DMSO, 1% polysorbate 80, and 94% saline to 1 mg/ml, and 10 mg/kg was injected intraperitoneally twice per day, or added This work was funded by grants from the Beta Cell Biology Consortium, JDRF, to cultured islets (10 mM). ERC, EU/FP7 (BetaCellTherapy, grant 241883), the Helmsley trust, the Dutch As controls for mice expressing mutant GCK, we used single transgenic friends of Hebrew University, and the I-CORE Program of The Israel Science littermates, which were identical to wild-type in terms of stress marker expres- Foundation #41.11 (to Y.D.). This work was supported in part by a grant sion and b cell apoptosis (not shown). from USAID’s American Schools and Hospitals Abroad Program for the up- Islets were isolated following collagenase perfusion and handpicked as grading of the Hebrew University Medical School Flow Cytometry laboratory. described (Porat et al., 2011). This study was performed with the support of the Network for Pancreatic The joint Hebrew University and Hadassah Medical Center ethics committee Organ Donors with Diabetes (nPOD), a JDRF-sponsored collaborative T1D (IACUC) approved the study protocol for animal welfare. The Hebrew Univer- research project. Organ Procurement Organizations (OPO) partnering with sity is AAALAC accredited. nPOD are listed at www.jdrfnpod.org/our-partners.php. O.Z. is a New York Stem Cell Foundation-Druckenmiller Fellow. M.S.-R. was supported by a post- doctoral fellowship from Novo Nordisk and A.H. by a fellowship from Mr. Alan Human Material Harper. We thank Norma Kidess-Bassir for excellent technical assistance, For the data presented in Figure 7, we used paraffin sections obtained from Donald K. Scott for the gift of rat glucokinase cDNA, Orr Spiegel for advice pancreata of brain-dead patients after Institutional Review Board (IRB) with statistical analysis, and Christian Broberger, Muli Ben-Sasson, and Ittai approval. Maximal warm ischemia time before fixation was 6 hr. Patient details Ben-Porath for insightful discussions. are presented in Table S1. Material from GCK-CHI and KATP-CHI patients (Figures 4 and S1) was Received: August 24, 2012 obtained after pancreatectomy as described (Kassem et al., 2000, 2010; Revised: March 30, 2013 Sayed et al., 2009; Snider et al., 2013). Age-matched controls shown in Accepted: November 6, 2013 Figure S1 were obtained from the Network for Pancreatic Organ Donors Published: December 12, 2013 (nPOD).

REFERENCES Histology and Immunostaining Tissue processing was performed as described (Nir et al., 2007) using Beltrand, J., Caquard, M., Arnoux, J.B., Laborde, K., Velho, G., Verkarre, V., antibodies listed in Table S2. TUNEL staining was performed using a kit Rahier, J., Brunelle, F., Nihoul-Fe´ ke´ te´ , C., Saudubray, J.M., et al. (2012). (In Situ Cell Death Detection Kit, AP; Roche). Fluorescent images were taken Glucose metabolism in 105 children and adolescents after pancreatectomy on a Nikon C1 confocal microscope at 4003 magnification. for congenital hyperinsulinism. Diabetes Care 35, 198–203. For b cell area determination, consecutive paraffin sections 75 mm apart, spanning the entire pancreas, were stained for insulin and hematoxylin. Digital Bensellam, M., Laybutt, D.R., and Jonas, J.C. (2012). The molecular mecha- 403 images were obtained, stitched together using NIS-Elements software, nisms of pancreatic b-cell glucotoxicity: recent findings and future research and the fraction of tissue stained for insulin was determined. directions. Mol. Cell. Endocrinol. 364, 1–27. For in vitro staining, islets were trypsinized and cytospun at 14,000 rpm for Butler, A.E., Janson, J., Bonner-Weir, S., Ritzel, R., Rizza, R.A., and Butler, 5 min. Slides were fixed for 10 min with paraformaldehyde, permeabilized for P.C. (2003). Beta-cell deficit and increased beta-cell apoptosis in humans 15 min, and then stained for insulin and gH2AX. with type 2 diabetes. Diabetes 52, 102–110.

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 119 Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

Cadet, J., Ravanat, J.L., TavernaPorro, M., Menoni, H., and Angelov, D. (2012). Karunakaran, U., Kim, H.-J., Kim, J.-Y., and Lee, I.-K. (2012). Guards and cul- Oxidatively generated complex DNA damage: tandem and clustered lesions. prits in the endoplasmic reticulum: glucolipotoxicity and b-cell failure in type II Cancer Lett. 327, 5–15. diabetes. Exp. Diabetes Res. 2012, 639762. Chang-Chen, K.J., Mullur, R., and Bernal-Mizrachi, E. (2008). Beta-cell failure Kassem, S.A., Ariel, I., Thornton, P.S., Scheimberg, I., and Glaser, B. (2000). as a complication of diabetes. Rev. Endocr. Metab. Disord. 9, 329–343. Beta-cell proliferation and apoptosis in the developing normal human Cuesta-Mun˜ oz, A.L., Huopio, H., Otonkoski, T., Gomez-Zumaquero, J.M., pancreas and in hyperinsulinism of infancy. Diabetes 49, 1325–1333. Na¨ nto¨ -Salonen, K., Rahier, J., Lo´ pez-Enriquez, S., Garcı´a-Gimeno, M.A., Kassem, S., Bhandari, S., Rodrı´guez-Bada, P., Motaghedi, R., Heyman, M., Sanz, P., Soriguer, F.C., and Laakso, M. (2004). Severe persistent hyperinsu- Garcı´a-Gimeno, M.A., Cobo-Vuilleumier, N., Sanz, P., Maclaren, N.K., linemic hypoglycemia due to a de novo glucokinase mutation. Diabetes 53, Rahier, J., et al. (2010). Large islets, beta-cell proliferation, and a glucokinase 2164–2168. mutation. N. Engl. J. Med. 362, 1348–1350. de Vathaire, F., El-Fayech, C., Ben Ayed, F.F., Haddy, N., Guibout, C., Winter, Kiyosue, A., Hayashi, N., Komori, H., Leonsson-Zachrisson, M., and D., Thomas-Teinturier, C., Veres, C., Jackson, A., Pacquement, H., et al. Johnsson, E. (2013). Dose-ranging study with the glucokinase activator (2012). Radiation dose to the pancreas and risk of diabetes mellitus in child- AZD1656 as monotherapy in Japanese patients with type 2 diabetes mellitus. hood cancer survivors: a retrospective cohort study. Lancet Oncol. 13, Diabetes Obes. Metab. 15, 923–930. 1002–1010. Las, G., and Shirihai, O.S. (2010). The role of autophagy in b-cell lipotoxicity Doliba, N.M., Qin, W., Najafi, H., Liu, C., Buettger, C.W., Sotiris, J., Collins, and type 2 diabetes. Diabetes Obes. Metab. 12 (Suppl 2 ), 15–19. H.W., Li, C., Stanley, C.A., Wilson, D.F., et al. (2012). Glucokinase activation Leibowitz, G., Glaser, B., Higazi, A.A., Salameh, M., Cerasi, E., and Landau, H. repairs defective bioenergetics of islets of Langerhans isolated from type 2 (1995). Hyperinsulinemic hypoglycemia of infancy (nesidioblastosis) in clinical diabetics. Am. J. Physiol. Endocrinol. Metab. 302, E87–E102. remission: high incidence of diabetes mellitus and persistent b-cell dysfunc- Eizirik, D.L., Cardozo, A.K., and Cnop, M. (2008). The role for endoplasmic re- tion at long-term follow-up. J. Clin. Endocrinol. Metab. 80, 386–392. ticulum stress in diabetes mellitus. Endocr. Rev. 29, 42–61. Maedler, K., Størling, J., Sturis, J., Zuellig, R.A., Spinas, G.A., Arkhammar, Gannon, M., Shiota, C., Postic, C., Wright, C.V., and Magnuson, M. (2000). P.O., Mandrup-Poulsen, T., and Donath, M.Y. (2004). Glucose- and Analysis of the Cre-mediated recombination driven by rat insulin promoter in interleukin-1beta-induced beta-cell apoptosis requires Ca2+ influx and extra- embryonic and adult mouse pancreas. Genesis 26, 139–142. cellular signal-regulated kinase (ERK) 1/2 activation and is prevented by a sul- fonylurea receptor 1/inwardly rectifying K+ channel 6.2 (SUR/Kir6.2) selective Girard, C.A., Wunderlich, F.T., Shimomura, K., Collins, S., Kaizik, S., Proks, P., potassium channel opener in human islets. Diabetes 53, 1706–1713. Abdulkader, F., Clark, A., Ball, V., Zubcevic, L., et al. (2009). Expression of an Marino, S., Vooijs, M., van Der Gulden, H., Jonkers, J., and Berns, A. (2000). activating mutation in the gene encoding the KATP channel subunit Kir6.2 in Induction of medulloblastomas in p53-null mutant mice by somatic inactivation mouse pancreatic beta cells recapitulates neonatal diabetes. J. Clin. Invest. of Rb in the external granular layer cells of the cerebellum. Genes Dev. 14, 994– 119, 80–90. 1004. Glaser, B., Kesavan, P., Heyman, M., Davis, E., Cuesta, A., Buchs, A., Stanley, Marselli, L., Thorne, J., Dahiya, S., Sgroi, D.C., Sharma, A., Bonner-Weir, S., C.A., Thornton, P.S., Permutt, M.A., Matschinsky, F.M., and Herold, K.C. Marchetti, P., and Weir, G.C. (2010). Gene expression profiles of Beta-cell (1998). Familial hyperinsulinism caused by an activating glucokinase mutation. enriched tissue obtained by laser capture microdissection from subjects N. Engl. J. Med. 338, 226–230. with type 2 diabetes. PLoS ONE 5, e11499. Glaser, B., Ryan, F., Donath, M., Landau, H., Stanley, C.A., Baker, L., Barton, Meacham, L.R., Sklar, C.A., Li, S., Liu, Q., Gimpel, N., Yasui, Y., Whitton, J.A., D.E., and Thornton, P.S. (1999). Hyperinsulinism caused by paternal-specific Stovall, M., Robison, L.L., and Oeffinger, K.C. (2009). Diabetes mellitus in long- inheritance of a recessive mutation in the sulfonylurea-receptor gene. term survivors of childhood cancer. Increased risk associated with radiation Diabetes 48, 1652–1657. therapy: a report for the childhood cancer survivor study. Arch. Intern. Med. Greig, N.H., Mattson, M.P., Perry, T., Chan, S.L., Giordano, T., Sambamurti, 169, 1381–1388. K., Rogers, J.T., Ovadia, H., and Lahiri, D.K. (2004). New therapeutic strategies Meininger, G.E., Scott, R., Alba, M., Shentu, Y., Luo, E., Amin, H., Davies, M.J., and drug candidates for neurodegenerative diseases: p53 and TNF-alpha in- Kaufman, K.D., and Goldstein, B.J. (2011). Effects of MK-0941, a novel gluco- hibitors, and GLP-1 receptor agonists. Ann. N Y Acad. Sci. 1035, 290–315. kinase activator, on glycemic control in insulin-treated patients with type 2 dia- Gu, G., Dubauskaite, J., and Melton, D.A. (2002). Direct evidence for the betes. Diabetes Care 34, 2560–2566. pancreatic lineage: NGN3+ cells are islet progenitors and are distinct from Miki, T., Iwanaga, T., Nagashima, K., Ihara, Y., and Seino, S. (2001). Roles of duct progenitors. Development 129, 2447–2457. ATP-sensitive K+ channels in cell survival and differentiation in the endocrine Guo, S., Dai, C., Guo, M., Taylor, B., Harmon, J.S., Sander, M., Robertson, pancreas. Diabetes 50 (Suppl 1 ), S48–S51. R.P., Powers, A.C., and Stein, R. (2013). Inactivation of specific b cell transcrip- Miller, F.D., Pozniak, C.D., and Walsh, G.S. (2000). Neuronal life and death: an tion factors in type 2 diabetes. J. Clin. Invest. Published online July 1, 2013. essential role for the p53 family. Cell Death Differ. 7, 880–888. http://dx.doi.org/10.1172/JCI65390. Moore, G., Knoblaugh, S., Gollahon, K., Rabinovitch, P., and Ladiges, W. Halazonetis, T.D., Gorgoulis, V.G., and Bartek, J. (2008). An oncogene- (2008). Hyperinsulinemia and in Wrn null mice fed a diabeto- induced DNA damage model for cancer development. Science 319, 1352– genic diet. Mech. Ageing Dev. 129, 201–206. 1355. Murtaugh, L.C., Stanger, B.Z., Kwan, K.M., and Melton, D.A. (2003). Notch Hinault, C., Kawamori, D., Liew, C.W., Maier, B., Hu, J., Keller, S.R., Mirmira, signaling controls multiple steps of pancreatic differentiation. Proc. Natl. R.G., Scrable, H., and Kulkarni, R.N. (2011). D40 Isoform of p53 controls Acad. Sci. USA 100, 14920–14925. b-cell proliferation and glucose homeostasis in mice. Diabetes 60, 1210–1222. Nakamura, A., Terauchi, Y., Ohyama, S., Kubota, J., Shimazaki, H., Nambu, T., Huopio, H., Otonkoski, T., Vauhkonen, I., Reimann, F., Ashcroft, F.M., and Takamoto, I., Kubota, N., Eiki, J., Yoshioka, N., et al. (2009). Impact of small- Laakso, M. (2003). A new subtype of autosomal dominant diabetes attributable molecule glucokinase activator on glucose metabolism and beta-cell mass. to a mutation in the gene for sulfonylurea receptor 1. Lancet 361, 301–307. Endocrinology 150, 1147–1154. James, C., Kapoor, R.R., Ismail, D., and Hussain, K. (2009). The genetic basis Nakamura, A., Togashi, Y., Orime, K., Sato, K., Shirakawa, J., Ohsugi, M., of congenital hyperinsulinism. J. Med. Genet. 46, 289–299. Kubota, N., Kadowaki, T., and Terauchi, Y. (2012). Control of beta cell function Kaneto, H., Miyatsuka, T., Kawamori, D., Yamamoto, K., Kato, K., Shiraiwa, T., and proliferation in mice stimulated by small-molecule glucokinase activator Katakami, N., Yamasaki, Y., Matsuhisa, M., and Matsuoka, T.A. (2008). PDX-1 under various conditions. Diabetologia 55, 1745–1754. and MafA play a crucial role in pancreatic beta-cell differentiation and mainte- Neema, M., Navarro-Quiroga, I., Chechlacz, M., Gilliams-Francis, K., Liu, J., nance of mature beta-cell function. Endocr. J. 55, 235–252. Lamonica, K., Lin, S.L., and Naegele, J.R. (2005). DNA damage and

120 Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. Cell Metabolism DNA Damage and p53 in b Cell Glucotoxicity

nonhomologous end joining in excitotoxicity: neuroprotective role of DNA- Shimoda, M., Kanda, Y., Hamamoto, S., Tawaramoto, K., Hashiramoto, M., PKcs in kainic acid-induced seizures. Hippocampus 15, 1057–1071. Matsuki, M., and Kaku, K. (2011). The human glucagon-like peptide-1 Nir, T., Melton, D.A., and Dor, Y. (2007). Recovery from diabetes in mice by analogue liraglutide preserves pancreatic beta cells via regulation of cell beta cell regeneration. J. Clin. Invest. 117, 2553–2561. kinetics and suppression of oxidative and endoplasmic reticulum stress in a mouse model of diabetes. Diabetologia 54, 1098–1108. Nito, C., Ueda, M., Inaba, T., Katsura, K., and Katayama, Y. (2011). FK506 ameliorates oxidative damage and protects rat brain following transient focal Shirakawa, J., Togashi, Y., Sakamoto, E., Kaji, M., Tajima, K., Orime, K., Inoue, cerebral ischemia. Neurol. Res. 33, 881–889. H., Kubota, N., Kadowaki, T., and Terauchi, Y. (2013). Glucokinase activation Nyblom, H.K., Bugliani, M., Fung, E., Boggi, U., Zubarev, R., Marchetti, P., and ameliorates ER stress-induced apoptosis in pancreatic b-cells. Diabetes 62, Bergsten, P. (2009). Apoptotic, regenerative, and immune-related signaling in 3448–3458. human islets from type 2 diabetes individuals. J. Proteome Res. 8, 5650–5656. Snider, K.E., Becker, S., Boyajian, L., Shyng, S.L., MacMullen, C., Hughes, N., Poitout, V., and Robertson, R.P. (2008). Glucolipotoxicity: fuel excess and Ganapathy, K., Bhatti, T., Stanley, C.A., and Ganguly, A. (2013). Genotype and beta-cell dysfunction. Endocr. Rev. 29, 351–366. phenotype correlations in 417 children with congenital hyperinsulinism. J. Clin. Porat, S., Weinberg-Corem, N., Tornovsky-Babaey, S., Schyr-Ben-Haroush, Endocrinol. Metab. 98, E355–E363. R., Hija, A., Stolovich-Rain, M., Dadon, D., Granot, Z., Ben-Hur, V., White, P., et al. (2011). Control of pancreatic b cell regeneration by glucose meta- Srinivas, S., Watanabe, T., Lin, C.S., William, C.M., Tanabe, Y., Jessell, T.M., bolism. Cell Metab. 13, 440–449. and Costantini, F. (2001). Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus. BMC Dev. Biol. 1,4. Rahier, J., Guiot, Y., and Sempoux, C. (2011). Morphologic analysis of focal and diffuse forms of congenital hyperinsulinism. Semin. Pediatr. Surg. 20, Suberbielle, E., Sanchez, P.E., Kravitz, A.V., Wang, X., Ho, K., Eilertson, K., 3–12. Devidze, N., Kreitzer, A.C., and Mucke, L. (2013). Physiologic brain activity Ritzel, R.A., Jayasinghe, S., Hansen, J.B., Sturis, J., Langen, R., and Butler, causes DNA double-strand breaks in neurons, with exacerbation by amy- P.C. (2010). Beta-cell selective K(ATP)-channel activation protects beta-cells loid-b. Nat. Neurosci. 16, 613–621. and human islets from human islet amyloid polypeptide induced toxicity. Wang, Y., Gao, L., Li, Y., Chen, H., and Sun, Z. (2011). Nifedipine Protects Regul. Pept. 165, 158–162. INS-1 b-Cell from High Glucose-Induced ER Stress and Apoptosis. Int. J. Rogakou, E.P., Pilch, D.R., Orr, A.H., Ivanova, V.S., and Bonner, W.M. (1998). Mol. Sci. 12, 7569–7580. DNA double-stranded breaks induce histone H2AX phosphorylation on serine 139. J. Biol. Chem. 273, 5858–5868. Weir, G.C., Laybutt, D.R., Kaneto, H., Bonner-Weir, S., and Sharma, A. (2001). Sakhi, S., Bruce, A., Sun, N., Tocco, G., Baudry, M., and Schreiber, S.S. (1994). Beta-cell adaptation and decompensation during the progression of diabetes. p53 induction is associated with neuronal damage in the central nervous sys- Diabetes 50 (Suppl 1 ), S154–S159. tem. Proc. Natl. Acad. Sci. USA 91, 7525–7529. Wicksteed, B., Brissova, M., Yan, W., Opland, D.M., Plank, J.L., Reinert, R.B., Sakuraba, H., Mizukami, H., Yagihashi, N., Wada, R., Hanyu, C., and Dickson, L.M., Tamarina, N.A., Philipson, L.H., Shostak, A., et al. (2010). Yagihashi, S. (2002). Reduced beta-cell mass and expression of oxidative Conditional gene targeting in mouse pancreatic ß-Cells: analysis of ectopic stress-related DNA damage in the islet of Japanese Type II diabetic patients. Cre transgene expression in the brain. Diabetes 59, 3090–3098. Diabetologia 45, 85–96. Wilding, J.P., Leonsson-Zachrisson, M., Wessman, C., and Johnsson, E. Sayed, S., Langdon, D.R., Odili, S., Chen, P., Buettger, C., Schiffman, A.B., (2013). Dose-ranging study with the glucokinase activator AZD1656 in patients Suchi, M., Taub, R., Grimsby, J., Matschinsky, F.M., and Stanley, C.A. with type 2 diabetes mellitus on metformin. Diabetes Obes. Metab. 15, (2009). Extremes of clinical and enzymatic phenotypes in children with hyper- 750–759. insulinism caused by glucokinase activating mutations. Diabetes 58, 1419– 1427. Yusta, B., Baggio, L.L., Estall, J.L., Koehler, J.A., Holland, D.P., Li, H., Schultz, L.B., Chehab, N.H., Malikzay, A., and Halazonetis, T.D. (2000). p53 Pipeleers, D., Ling, Z., and Drucker, D.J. (2006). GLP-1 receptor activation binding protein 1 (53BP1) is an early participant in the cellular response to improves beta cell function and survival following induction of endoplasmic DNA double-strand breaks. J. Cell Biol. 151, 1381–1390. reticulum stress. Cell Metab. 4, 391–406.

Cell Metabolism 19, 109–121, January 7, 2014 ª2014 Elsevier Inc. 121