Beta cell adaptation to pregnancy requires prolactin action on both beta and non-beta cells

Pancreatic islets adapt to insulin resistance of pregnancy by up regulating β-cell mass and increasing insulin secretion. Previously, using a transgenic mouse with global, heterozygous deletion of prolactin receptor (Prlr+/−), we found Prlr signaling is important for this adaptation. However, since Prlr is expressed in tissues outside of islets as well as within islets and prolactin signaling affects β-cell development, to understand β-cell-specific effect of prolactin signaling in pregnancy, we generated a transgenic mouse with an inducible conditional deletion of Prlr from β-cells. Here, we found that β-cell-specific Prlr reduction in adult mice led to elevated blood glucose, lowed β-cell mass and blunted in vivo glucose-stimulated insulin secretion during pregnancy. When we compared gene expression profile of islets from transgenic mice with global (Prlr+/−) versus β-cell-specific Prlr reduction (βPrlR+/−), we found 95 differentially expressed gene, most of them down regulated in the Prlr+/− mice in comparison to the βPrlR+/− mice, and many of these genes regulate apoptosis, synaptic vesicle function and neuronal development. Importantly, we found that islets from pregnant Prlr+/− mice are more vulnerable to glucolipotoxicity-induced apoptosis than islets from pregnant βPrlR+/− mice. These observations suggest that down regulation of prolactin action during pregnancy in non-β-cells secondarily and negatively affect β-cell gene expression, and increased β-cell susceptibility to external insults.

Pregnancy is characterized by insulin resistance, to shunt nutrients from the mother to the developing fetus. Maternal pancreatic β-cells adapt to the insulin resistance of pregnancy by up regulating islet mass and insulin secretion, a process that quickly reverses at parturition [1][2][3] . In animal studies, pregnancy-induced β-cell adaptation consists of (1) a lower threshold for glucose-stimulated insulin secretion, (2) a higher insulin content, and (3) a higher β-cell proliferation rate [4][5][6][7] . Both in vitro and in vivo observations support a role for prolactin (PRL) and/or placental lactogens (PLs) in this adaptive process. First, prolactin receptor (Prlr), the receptor for both PRL and PLs, is present on pancreatic β-cells, and Prlr expression increases during pregnancy 3,8,9 . Second, the rise in the levels of PRL and PLs parallels the increases in β-cell mass and glucose-stimulated insulin secretion during pregnancy 3 . Third, in vitro exposure of isolated islets to PRL/PLs increases insulin secretion and β-cell proliferation, and lowers the threshold of glucose-stimulated insulin secretion 10,11 , mimicking the effects of pregnancy on β-cells. To determine whether Prlr signaling is required for β-cell adaptation to pregnancy in vivo, transgenic mice with global or β-cell-specific Prlr knockout have been examined. Using heterozygous global prolactin receptor-null mice (Prlr +/− ), we established the important in vivo role of Prlr on β-cell adaptation to pregnancy 12 . We found that in comparison to their wild type littermates, pregnant Prlr +/− dams were glucose intolerant and secreted less insulin in response to an intraperitoneal glucose tolerance test (IPGTT). They also had a lower β-cell proliferation rate and lower β-cell mass than the wild type (i.e. Prlr +/+ ) mice 12 . We could not use the homozygous Prlr −/− mice because they have a placental implantation defect, thus are infertile 13 . Two separate groups have since generated β-cell-specific Prlr-knockout mice, one using the rat insulin promoter (RIP) controlled Cre (i.e. RIP-Cre) 14 and the other using the pancreatic and duodenal homeobox 1 (Pdx1) promoter controlled Cre (i.e. Pdx1-Cre) 15 . Similar to our findings in the Prlr +/− mice, β-cell-specific Prlr-deletion caused gestational diabetes, accompanied by a reduction in β-cell proliferation and a blunted glucose-stimulated insulin secretion in vivo, which is largely due to the lower β-cell mass rather than an intrinsic insulin secretion defect Gestational glucose intolerance in βPrlR +/− mice. Next, we determined glucose homeostasis of the βPrlR +/− mice during pregnancy. Intraperitoneal glucose tolerance test (IPGTT) was performed on day 15 of pregnancy. In comparison to their wild type littermates (i.e. βPrlR +/+ mice), the βPrlR +/− mice were glucose intolerant with a greater glucose excursion, expressed as integrated area under the curve over the 120 min of the IPGTT (βPrlR +/− : AUC = 1770 ± 85.21 mM × min vs. βPrlR +/+ : AUC = 1325 ± 120.2 mM × min, n = 10-16) ( Fig. 2A,B). Interestingly, transgenic mice heterozygous for global prolactin receptor deletion, Prlr +/− , had even higher blood glucose at 45 and 60 min of the IPGTT than the βPrlR +/− mice. Non-fasted blood glucose on day 14 of pregnancy is also higher in the βPrlR +/− mice than their wild type littermates (8.09 ± 0.30 mM vs. 7.0 ± 0.24 mM for βPrlR +/+ mice) (Fig. 2C), and it is higher still in the Prlr +/− mice (9.17 ± 0.29 mM), although there is no significant difference in fasting blood glucose amongst the 3 genotypes (data not shown). Of note, the presence of Cre recombinase and the administration of tamoxifen per se had no effect on glucose homeostasis, as the βPrlR +/+ and the floxed-PrlR mice (which do not have Cre recombinase and never received tamoxifen) had comparable glucose excursion during an IPGTT, with AUC of 1325 ± 120.2 mM × min and 1325.3 ± 261.9 mM × min, respectively. We did not observe a significant difference in glucose response during an insulin tolerance test amongst the βPrlR +/+ , βPrlR +/− , and the Prlr +/− mice on day 14 of pregnancy ( Fig. 2D); therefore, the difference in glucose tolerance observed was not a result of difference in insulin sensitivities. βPrlR +/− mice had lower β-cell mass and secreted less insulin. Similar to our previous observation in the heterozygous global Prlr-deletion mice (Prlr +/− ) 12 , βPrlR +/− mice had a lower β-cell mass in comparison to their wild type littermates during pregnancy (Fig. 3A). During an IPGTT, both the βPrlR +/− and Prlr +/− mice secreted less insulin in comparison to the wild type mice (Fig. 3B). This blunted in vivo glucose-stimulated insulin secretion is not likely due to an intrinsic insulin secretory defect because we found that islets from βPrlR +/+ and βPrlR +/− mice were equally responsive to glucose in vitro, suggesting that the difference in serum insulin levels during an IPGTT likely reflects the smaller β-cell mass found in the mutant mice.
Prolactin receptor has non-cell-autonomous effects on β-cell gene expression during pregnancy. The above results demonstrated that transgenic mice with global (Prlr +/− ) or β-cell-specific prolactin receptor deletion (βPrlR +/− ) are phenotypically similar in the metabolic parameters measured: in comparison to their wild type littermates, both mutant mice are glucose intolerant on day 15 of pregnancy, with a lower β-cell mass, which leads to a lower amount of insulin being secreted in response to a glucose load, supporting the functional importance of prolactin receptor in islet adaptation to pregnancy. To understand the effect of Prlr action in non-β-cells vis-à-vis β-cells in regulating β-cell adaptation to pregnancy, we compared gene expression profile of islets from Prlr +/− to that of βPrlR +/− mice. It is important to note that in Prlr +/− mice, the reduction in Prlr expression would be widespread, including β-cells as well as non-β-cells within the islets such as endothelial cells that lines the vasculature and neurons that innervates the islets. Furthermore, Prlr expression is also reduced in tissues outside of islets, including but not limited to tissues such as the hypothalamus, muscle, fat, and liver 20 . Using Sleuth 21 and a false discovery rate (FDR) threshold of 0.05, we identified 95 differentially expressed genes in islets (Table 1 and Fig. 4); most of them are down regulated in the Prlr +/− mice in comparison to the βPrlR +/− mice, with estrogen receptor 1 (ESR1) being the most significantly affected gene (Fig. 5A). Since estrogen receptor is known to be expressed in β-cells in islets and activation of estrogen receptor has been shown www.nature.com/scientificreports/ to protect β-cells from apoptosis 22 , we compared the susceptibility of islets isolated from the pregnant Prlr +/− and βPrlR +/− mice to fatty-acid-mediated induction of apoptosis in vitro [22][23][24] . We treated isolated islets with 0.5 mM of palmitate or BSA control for 72 h. Using cleaved caspase-3 positivity as a marker of activation of the apoptosis  Supplementary Fig. S1. (C) Image of pancreas from a βPrlR −/− mouse expressing mT/mG, which allows detection of Cre recombination, visualized as green fluorescence (arrows). The areas of green fluorescence represent islets. (D) Relative mRNA expression of prolactin receptor in islets, liver, fat, and hypothalamus. n = 3-12 mice/genotype. "*" = p < 0.05 and "**" = p < 0.01 in comparison to the βPrlR +/+ mice. Prlr +/− and Prlr +/+ denote transgenic mice with a global, heterozygous deletion of prolactin receptor or their wild type littermates, respectively. βPrlR −/− and βPrlR +/− denote transgenic mice with a β-cell-specific homozygous or heterozygous prolactin receptor deletion, while βPrlR +/+ denotes their wild type littermates. www.nature.com/scientificreports/  www.nature.com/scientificreports/ pathway, we found that palmitate induced a greater increase in the percentage of insulin-positive cells that are also positive for cleaved caspase-3 in islets from Prlr +/− mice in comparison to islets from βPrlR +/− mice (Fig. 5B). This is consistent with the pro-survival role of the estrogen receptor, which is expressed at a lower level in islets of pregnant Prlr +/− mice than in βPrlR +/− mice. This suggests that the down regulation of Prlr action in non-β-cells within and outside islets, as would be expected in the Prlr +/− mice, increased β-cell susceptibility to apoptosis. Next, using Ingenuity Pathway Analysis, we identified several canonical pathways that are differentially expressed, including citrulline cycle and metabolism, dopamine receptor signaling, synaptogenesis signaling pathway and the serotonin receptor-signaling pathway, the latter has previously been found to regulate β-cell proliferation during pregnancy 25 (Table 2). KEGG pathway analysis of the 95 genes identified enrichment of genes involved in cholinergic synapse (p = 3.39 × 10 -4 ) and synaptic vesicle cycle (p = 4.96 × 10 -3 ) ( Table 2). GO term analysis found enrichment in genes that are involved in synaptic transmission (p = 6.45 × 10 -4 ), nervous system development (p = 6.03 × 10 -4 ), anterograde trans-synaptic signaling (p = 3.62 × 10 -3 ), noradrenergic neuron differentiation (p = 4.57 × 10 -3 ), and cell differentiation in hindbrain (p = 2.26 × 10 -2 ) ( Table 2). Using RT-qPCR, we confirmed that acetylcholinesterase (Ache) and synaptotagmin 1 (Syt1), both identified in the KEGG pathway and GO term analysis (Table 2), expressed at a lower level in islets of pregnant PrlR +/− mice in comparison to the βPrlR +/− mice. In addition, BEN domain containing 5 (Bend5), a transcription repressor, cyclin G1 (Ccng1), which regulates cell cycle and growth, and ELAV like RNA binding protein 3 (Elavl3), a neural-specific www.nature.com/scientificreports/ RNA-binding protein, also expressed at a lower level in islets from pregnant Prlr +/− mice in comparison to the βPrlR +/− mice (Fig. 5A). These differences in gene expression between the pregnant Prlr +/− and βPrlR +/− were not due to difference in developmental expression prior to pregnancy because except for Bend5, we did not detect a significant difference in expression of these genes between non-pregnant Prlr +/− and βPrlR +/− mice. One common theme from these pathway analyses is the down regulation of genes that regulate neuronal structure and function as a result of global down regulation of prolactin receptor. This is highly interesting because previously, prolactin receptor action was shown to be important in neurogenesis in the forebrain subventricular zone of female mice during pregnancy, an area of the brain that regulates maternal behavior 26 .  www.nature.com/scientificreports/

Discussion
Pregnancy is an insulin resistant state and maternal pancreatic islets adapt to the increase in insulin demand by up regulating β-cell mass and increasing insulin secretion. Previously, we showed that a transgenic mouse with a global, heterozygous deletion of prolactin receptor (Prlr +/− ) was glucose intolerant during pregnancy. This was accompanied by a reduction in β-cell mass and a blunted glucose-stimulated insulin secretion in vivo, commensurate with the lower β-cell mass 12 . A caveat of using a global knockout mouse is that the prolactin receptor is widely expressed 18,27 , raising the possibility that the phenotype observed in Prlr +/− mice may be secondary to prolactin action in cells other than β-cells. To address this possibility, we generated a transgenic mouse with an inducible, conditional prolactin receptor gene knockout. We chose to make the gene deletion inducible because β-cell-specific promoters such as the insulin promoter or the Pdx1 promoter are active during embryonic development of the pancreas 28 . In fact, Auffret et al. found that transgenic mice with a global homozygous deletion of prolactin receptor (i.e. Prlr −/− ) have abnormal β-cell mass expansion both embryonically and during perinatal life 16 . They found a 30% reduction in β-cell mass in the Prlr −/− newborns. A transgenic mouse with an inducible promoter such that the prolactin receptor gene is only deleted during adulthood would circumvent this developmental defect and allow us to study prolactin receptor's effect specifically during pregnancy. Here, we report that in βPrlR +/− mice, the deletion of prolactin receptor in adulthood, before pregnancy, led to maladaptation to insulin resistance of pregnancy, with a lower β-cell mass and insulin secretion, as well as a higher blood glucose level in the βPrlR +/− mice in comparison to their wild type littermates. As expected, β-cell-specific deletion of prolactin receptor did not have a significant effect on overall insulin sensitivity, while the Prlr +/− mice had a tendency for slightly lower insulin sensitivity at 45 min of the ITT (Fig. 2D). This tendency to increased insulin sensitivity however, is not likely to be a significant driver of the lower β-cell mass and insulin secretion in the Prlr deficient mice because if the lower insulin secretion observed in the Prlr +/− mice reflected their improved insulin sensitivity thereby a reduced need for insulin 29 , we would anticipate Prlr +/− mice to have lower blood glucose than βPrlR +/− mice. Our results showed that both Prlr +/− and βPrlR +/− have higher blood glucose than the wild type mice and in fact, Prlr +/− mice had slightly higher, not lower blood glucose than βPrlR +/− mice. Therefore, the lower insulin sensitivity in the Prlr +/− mice at 45 min of ITT is not likely a significant determinant of β-cell mass and insulin secretion. Other groups have since generated conditional knockouts of the prolactin receptor gene. Banerjee et al. used a RIP-Cre promoter to generate a conditional Prlr knockout 14 . They found that β-cell-specific deletion of Prlr led to gestational diabetes due to reduced β-cell proliferation and a failure to expand β-cell mass during pregnancy. They identified MafB as one of the Prlr-signaling targets, and MafB deletion in maternal β-cells caused gestational diabetes as well. Subsequently, they performed a transcriptome analysis and found that forkhead box protein M1, polycomb repressor complex 2 subunits, Suz12 and enhancer of zeste homolog 2 are additional Prlr signaling targets 30 . Interestingly, Prlr was not required for β-cell adaptation to high fat feeding. In fact, pregnancy and high fat feeding activate very different genes in the islets, suggesting that the two metabolic stressors engage different mechanisms to adapt. Nteeba et al. also generated a β-cell-specific Prlr knockout mouse, using a Pdx1-Cre www.nature.com/scientificreports/ promoter 15 . In agreement with Banerjee et al., they found glucose intolerance during pregnancy. Interestingly, the β-cell-specific Prlr knockout mice had lower β-cell mass in the non-pregnant state and the pregnancy-stimulated β-cell mass expansion was essentially non-existent in their model. Moreover, they found that the loss of maternal pancreatic Prlr signaling was associated with fetal overgrowth and dysregulation of prolactin-associated genes in the placenta. The functional consequence of the placental findings was not explored in that study, but it is of high interest because we previously reported that Prlr deletion had deleterious, transgenerational effect such that the wild type offspring from Prlr +/− dams had maladaptive β-cell responses during pregnancy 31 , which could be secondary to an abnormal placenta. Prolactin receptor expression is widespread. Amongst endocrine cells of the pancreatic islet, Prlr expression is limited to β-cells but it is also expressed in endothelial cells that lines the blood vessels and neurons, both are non-β-cells that are present in islets. As well, Prlr is expressed in many non-β-cells outside of islets, such as in the brain, muscles, fat, and liver. To determine whether reduction of Prlr in non-β-cells, including both nonβ-cells within islets and those outside of the islets, has an effect on islet gene expression during pregnancy, we performed RNAseq analysis, comparing gene expression of islets of Prlr +/− versus βPrlR +/− mice on day 15 of pregnancy. Interestingly, 95 genes are differentially expressed, and 89 out of 95 are down regulated in the islets of Prlr +/− mice in comparison to βPrlR +/− mice, with estrogen receptor 1 (ESR1) being the most significantly down regulated gene. This is interesting because Le May et al. showed that ERα is expressed in β-cells and activation of estrogen receptor protects β-cells from apoptosis [22][23][24] . Our previous observation that islets from pregnant mice are more resistant to free fatty acid induced apoptosis than islets from non-pregnant mice 32 may be in part attributable to the > 20-fold induction of ESR1 expression in islets during pregnancy. Indeed, when we exposed isolated islets from pregnant Prlr +/− and βPrlR +/− mice to free fatty acid, we observed a greater increase in percentage of activated caspase-3 positive β-cells relative to untreated cells in the Prlr +/− islets in comparison to the βPrlR +/− islets. A potential link between Prlr signaling and ESR1 expression is TGFβ, as our previous work on Prlr +/− mice found that TGFβ is up regulated in Prlr +/− mice in comparison to Prlr +/+ mice, and TGFβ has been shown to regulate both Prlr and ESR1. Other potential factors that may act in an endocrine or a paracrine fashion to regulate β-cell gene expression include growth hormone [32][33][34] and hepatic growth factor 35 , as both are secreted from non-β-cells outside of the islets and expressed in non-β-cells within the islets, such as endothelial cells 32,36,37 , and both hormones have pro-proliferative and pro-survival effects on β-cells. Interestingly, growth hormone has been shown to regulate Syt1 expression 38 , another gene that is expressed at a lower level in the Prlr +/− mice in comparison to the βPrlR +/− mice (Fig. 5). It is important to note that since the RNAseq analysis was performed using whole islets and not purified β-cells, the difference in gene expression may reflect differential gene expression in various cell types in the islets and not only the difference in β-cells. While this is possible, we believe the difference in gene expression most likely reflects expression in β-cells because on day 15 of pregnancy, β-cells constitute 85-90% of endocrine cells in an islet and endothelial cells constitute ~ 10-15% of islet mass while neurons contribute minimally to islet mass (preliminary data not shown). Therefore, β-cells are the most abundant cell type in an islet during pregnancy and the gene expression changes are most likely occurring in β-cells. The β-cell gene expression difference between islets of βPrlR +/− and Prlr +/− mice however, are likely secondary to the down regulation of Prlr in non-β-cells because βPrlR +/− and Prlr +/− mice have comparable Prlr expression in islets (Fig. 1D) but Prlr +/− (and not βPrlr +/− ) mice also have down regulation of Prlr in non-β-cells, both within and outside of islets. This led us to speculate that there is Prlr-dependent endocrine (from nonβ-cells outside of islets) and/or paracrine factors (from non-β-cells within islets) that affect gene expression in β-cells in βPrlr +/− and Prlr +/− mice.
An interesting finding from our RNAseq analysis was the identification of serotonin receptor signaling as one of the top canonical pathways that are differentially regulated between Prlr +/− and βPrlR +/− mice 25,39 . Kim et al. reported that serotonin acts downstream of Prlr to stimulate β-cell proliferation, and this was through the Gαq-linked serotonin receptor 5-hydroxytryptamine receptor-2b (Htr2b) 25 . In a subsequent study, they found that another serotonin receptor, Htr3a, is important for regulation of insulin secretion 40 , although we did not detect a difference in Htr3a expression between the Prlr +/− and βPrlR +/− mice in this study.
Hence, observations from our β-cell-specific Prlr knockout mice as well as those from Banerjee et al. and Nteeba et al. 14,15 , and the global Prlr knockout mice 12 , indicate that Prlr has cell-autonomous effects on β-cell adaptation to pregnancy, which is accountable for most of the observed Prlr effect on β-cell mass expansion and insulin secretion during pregnancy. The current study reports the novel finding that the absence of Prlr action in non-β-cells increased β-cell's susceptibility to apoptosis in response to an external insult such as glucolipotoxicity, and that Prlr action in non-β-cells also affect β-cell gene expression during pregnancy, suggesting the presence of Prlr-dependent 'factor(s)' from non-β-cells within the islets and/or tissues outside of islets that regulate β-cell adaptation to pregnancy.

Methods
Ethical approval. All experimental procedures were approved by the Animal Use Review Committee at the University of Calgary in accordance with standards of the Canadian Council on Animal Care. The study was carried out in compliance with the ARRIVE guidelines.
Mice. Heterozygous prolactin receptor null mice (Prlr +/− ) on a C57BL/6 background were purchased from Jackson Laboratory and a working stock was generated by crossing Prlr +/− mice with wild type Prlr +/+ mice. The pups were genotyped as previously described 41 . Prlr +/+ and Prlr +/− mice did not receive tamoxifen.
Generation of the conditional prolactin-receptor null mice. A promoter-driven targeting cassette for the generation of a knockout-first allele with potential for conditional knockout was obtained from  43 , which has the Cre-ER gene under the control of the Pdx1 promoter, to generate the Pdx1CreER:PrlR +/− mice. Pdx1 is expressed in the pancreatic bud during embryonic development and it is required for the formation of pancreas; in adult pancreas, its expression is restricted to the β cells 44 . The male Pdx1CreER:PrlR +/− mice were crossed with female PrlR +/− mice to generate the homozygous conditional knockout Pdx1CreER:PrlR −/− (herein denoted as βPrlR −/− ), and the heterozygous conditional knockout βPrlR +/− , and the control littermates, i.e. Pdx1CreER, PrlR +/− and PrlR −/− mice. The βPrlR −/− mice were then crossed with the mT/mG reporter mice (The Jackson Laboratory, Stock No.: 007576), which carries ROSA mT/mG , a cell membrane-targeted, two-color fluorescent Cre-reporter allele. Before Cre recombination, the tdTomato (mT) fluorescence expression is widespread in all cells. Upon Cre recombinase activation, the cells express membrane-localized EGFP (mG) fluorescence, replacing the red fluorescence 45 . To induce Cre recombinase activity, tamoxifen (dissolved in corn oil) was given by oral gavage at age 8 weeks for 5 days at a dose of 4 mg/day (for a 20 g mouse). Mice were used 4 weeks after tamoxifen administration.
Mice were maintained on a 12-h light, 12-h dark cycle with liberal access to food and water. Mice were studied at 4-5 months of age. For pregnancy, female mice were set up with wild type male mice and the morning when a vaginal plug was found was designated as day 1 of pregnancy. Mice were used on day 15 of pregnancy. We chose to study day 15 of pregnancy because β-cell proliferation and β-cell mass peaks on days 14-15 of pregnancy 3,9 . Intraperitoneal glucose tolerance test (IPGTT) and insulin tolerance test (ITT). IPGTT (20% d-glucose solution, 2 g/kg body weight) and ITT (NovoRapid insulin, 0.75 unit/kg body weight) were performed as previously described 12 . Additional blood samples (~ 30 µl) were taken at times 0, 5, and 30 min of IPGTT for insulin concentration measurements by ELISA (ALPCO, catalog #80-INSMSU-E01). Non-fasted blood glucose was determined using a glucometer (FastTake) by sampling from tail vein at 8 a.m., and at the same time, an additional 30 μl of serum was taken from the saphenous vein and stored at − 80 °C for later measurement of insulin by ELISA.

Materials.
Collagenase P (Ref#11 213 865 001) was purchased from Roche. Trypsin 1 mg tablets (T7168) and all chemicals are from Sigma. Sources of antibodies are specified below.
Immunostaining. For quantification of β-cell mass, pancreas was isolated from pregnant and non-pregnant mice, weighed, embedded in paraffin blocks, longitudinally serial sectioned to 7 μm, then stained for insulin to identify β cells as previously described 12 . Briefly, after 1 h of blocking with 1% goat serum/phosphate buffered saline (PBS) at room temperature, tissues were incubated with primary antibody over night at 4 °C (guinea pig anti-insulin at 1:750, DAKO; diluted in 1% goat serum/PBS). This was followed by 1-h incubation with fluorophore-conjugated secondary antibodies (Cy3-anti-guinea pig at 1:300, Jackson ImmunoResearch, diluted in 1% goat serum/PBS). Bis-benzimide H 33342 trihydrochloride (0.1 μg/ml, Sigma) was added to the secondary antibody for nuclear staining. Stained sections were mounted using DakoCytomation fluorescent Mounting Medium (Agilent/Dako, Santa Clara, USA) and stored at 4 °C. To localize estrogen receptor-α (ERα) in pancreatic tissues, antigen retrieval was achieved by microwaving the paraffin embedded tissue sections in sodium citrate (pH 6), followed by 1 h of blocking with 1% goat serum/PBS at room temperature, then incubation with primary antibodies overnight at 4 °C to co-stain for insulin (to identify β-cells) and ERα (rabbit anti-ERα at 1:100, Santa Cruz). Next day, after 3 brief washes in PBS, tissue sections were incubated with secondary antibodies (Cy3-antirabbit at 1:300 and Alexa488-anti-guinea pig at 1:300, both diluted in 1% goat serum/PBS, plus Hoechst 33342 for nuclear staining) at room temperature for 1 h. For identification of cleaved caspase-3 positive cells, isolated mouse islets were incubated in 0.5 mM of palmitate in the presence of 33 mM glucose (see below) for 72 h. At the end of the incubation, islets were briefly washed in PBS, dissociated into single cells using 2 mM EDTA, and place on poly-l-lysine coated glass slides. The cells were fixed with 4% PFA in PBS for 10 min at room temperature and permeabilized with 0.2% Triton X for 15 min, then incubated with rabbit cleaved caspase-3 antibody (Promega, Madison, USA) at 1:1000 dilution in 1% goat serum/PBS at 4 °C overnight, followed by incubation with goat-anti-rabbit-Cy3 (Jackson ImmunoResearch, West Grove, USA) or goat-anti-rabbit-Alexa-488 antibodies (Molecular Probes, Waltham, USA) at 1:300 dilution in 1% goat serum/PBS at room temperature for 1 h. After 3 washes with PBS, cells were mounted with Dako fluorescent Mounting Medium.
Islet morphometry. Consecutive images of non-overlapping, adjacent areas of the entire pancreas section were acquired using a Zeiss fluorescence microscope, and captured with a CoolSnap digital camera 12 . Images were analyzed by ImageJ software (https:// imagej. nih. gov/ ij/) to measure the insulin-positive area as well as the area of the entire pancreas section (identified by nuclear staining). β-cell mass was calculated by multiplying the pancreas weight by the β-cell fraction (i.e. the ratio of insulin-positive cell area to total pancreatic tissue area www.nature.com/scientificreports/