Perk Gene Dosage Regulates Glucose Homeostasis by Modulating Pancreatic β-Cell Functions

Background Insulin synthesis and cell proliferation are under tight regulation in pancreatic β-cells to maintain glucose homeostasis. Dysfunction in either aspect leads to development of diabetes. PERK (EIF2AK3) loss of function mutations in humans and mice exhibit permanent neonatal diabetes that is characterized by insufficient β-cell mass and reduced proinsulin trafficking and insulin secretion. Unexpectedly, we found that Perk heterozygous mice displayed lower blood glucose levels. Methodology Longitudinal studies were conducted to assess serum glucose and insulin, intracellular insulin synthesis and storage, insulin secretion, and β-cell proliferation in Perk heterozygous mice. In addition, modulation of Perk dosage specifically in β-cells showed that the glucose homeostasis phenotype of Perk heterozygous mice is determined by reduced expression of PERK in the β-cells. Principal Findings We found that Perk heterozygous mice first exhibited enhanced insulin synthesis and secretion during neonatal and juvenile development followed by enhanced β-cell proliferation and a substantial increase in β-cell mass at the adult stage. These differences are not likely to entail the well-known function of PERK to regulate the ER stress response in cultured cells as several markers for ER stress were not differentially expressed in Perk heterozygous mice. Conclusions In addition to the essential functions of PERK in β-cells as revealed by severely diabetic phenotype in humans and mice completely deficient for PERK, reducing Perk gene expression by half showed that intermediate levels of PERK have a profound impact on β-cell functions and glucose homeostasis. These results suggest that an optimal level of PERK expression is necessary to balance several parameters of β-cell function and growth in order to achieve normoglycemia.


Introduction
The endocrine pancreatic b-cells have an exclusive and singular function to synthesize and secrete insulin. While insulin is essential for maintaining glucose homeostasis, hyperinsulinemia can result in hypoglycemic shock and death. Therefore insulin synthesis and secretion must be tightly regulated to provide the appropriate level of circulating insulin in response to episodic input of dietary carbohydrates and release of glucose stores. Pancreatic insulin output is controlled by a combination of regulating b-cell mass in the endocrine pancreas [1][2][3][4] and by regulating insulin synthesis and secretion in b-cells [5][6][7][8][9]. Although a large number of genes have been shown to influence b-cell growth and insulin synthesis and secretion, a small number of genes (ca. 20) including Perk have been identified in humans that are absolutely essential for b-cell growth or insulin production [10,11]. The consequence of the loss of function mutations in these genes is permanent neonatal diabetes (PND). Among these PND genes, the function of the Perk (EIF2AK3) gene has been the most controversial and perplexing [12][13][14][15]. Perk was initially identified as one of the three regulatory arms of the ER stress response pathway in cultured mammalian cells [16,17]. Shortly after its discovery [18] and characterization in cell culture, mutations in Perk were found to be the cause of the Wolcott-Rallison syndrome (WRS) in humans [19] that featured permanent neonatal diabetes, exocrine pancreas deficiency, growth retardation, and osteopenia. Perk knockout (KO) mouse strains were generated by us [15] and by Harding and Ron [12], which exhibited a nearly identical phenotype to that seen in human WRS patients, including permanent neonatal diabetes. By generating and analyzing tissue-specific Perk KO and transgenic rescue strains, we showed that the neonatal diabetes was caused by deficient b-cell growth and multiple problems in proinsulin synthesis and trafficking and insulin secretion [13,14,20].
An extensive analysis of PERK function by us has failed to support the initial hypothesis that the b-cell defects seen in Perk deficiency are due to misregulation of the ER stress response pathway [13,14]. Moreover, mutations in the other two regulatory arms of the ER stress pathway, ATF6 and IRE1, do not cause major b-cell dysfunctions or diabetes [21,22]. This demonstrates that dysfunction in the ER stress response generally does not result in permanent neonatal diabetes. Some of these b-cell dysfunctions seen in Perk KO mice can be attributed to the lack of phosphorylation of eIF2a, the primary substrate of PERK, because mutations that block the Ser51 phosphorylation site either in whole animals or in just the b-cells also result in diabetes [23,24]. However, other PERK-dependent b-cell functions may be independent of eIF2a phosphorylation including regulation of secretagogue stimulated calcium influx and insulin secretion [25].
Humans and mice that are heterozygous for a loss-of-function Perk mutation do not exhibit overt abnormal phenotypes [15,19,26]. However, we found that Perk heterozygous (Perk +/2 ) mice exhibit significantly lower serum blood glucose levels among several hundred litters of mice analyzed over the past ten years, opposite to the Perk KO mice which are severely hyperglycemic. To determine the underlying reasons for this shift in glucose homeostasis of Perk +/2 mice, we conducted a postnatal developmental analysis of b-cell growth and function in Perk +/2 mice compared to their homozygous wild-type littermates. We found that Perk +/2 mice first exhibited enhanced insulin synthesis and secretion during neonatal and juvenile development followed later at the adult stage by enhanced b-cell proliferation and a substantial increase in b-cell mass. These findings support the hypothesis that PERK dynamically regulates b-cell growth, insulin synthesis and secretion during postnatal development.

Genetic Strains
Perk global KO allele, floxed Perk allele were generated as previously described [15]. A Perk transgene under the control of the rat insulin promoter was introduced into the wild-type mice to generate mice with overexpression of Perk specifically in b-cells with an otherwise wild-type background (Perk +/+ ;bPerk) [27]. Perk +/2 mice that carried one Perk KO allele were in congenic C57Bl/6, 129 SvEvTac, or in a mixed background. To generate pancreatic specific Perk +/2 mice, pdx1-cre transgenic mice were crossed with mice homozygous for floxed Perk allele. Mice were sacrificed by CO 2 euthanasia. All animal studies were approved by the Institutional Animal Care and Use Committee of Pennsylvania State University, and all efforts were made to minimize suffering.

Cell culture
INS1 832/13 cells containing a short-hairpin RNA directed against the rat Perk mRNA (shPerk) were obtained from Dr. Fumihiko Urano (University of Massachusetts). The shPerk is stably integrated into the genome of INS1 832/13 b-cell lines and under the inducible regulation of doxycycline. The INS1 832/13 shPerk cells were cultured in a tetracycline-free environment to avoid leaky expression of shPerk.

Determination of serum glucose and glucose tolerance test
Blood samples were obtained from the tails and glucose were measured using OneTouch Ultra glucose meters. Glucose tolerance tests were performed on mice fasted 4 hours (for P17 mice) or overnight (for P50 mice) and injected intraperitoneally with 2 mg glucose/g of body weight.

Insulin measurement
Insulin concentrations were determined by immunoassay (Meso Scale Discovery, MSD). For serum insulin measurement, serum was obtained by centrifugation of blood samples at 10,000 g for 5 min. For islet and pancreatic insulin measurement, islets or pancreata were sonicated in 1 ml of cold acid ethanol (1.5% volume HCl in 75% ethanol). Insulin concentrations were further normalized to total protein concentration (determined by BIO-RAD Protein Assay). For studies of glucose stimulated insulin secretion, isolated islets or cultured b-cell line were firstly cultured overnight at 37uC (5% CO 2 ) in RPMI1640 medium containing 10% fetal bovine serum and 5.5 mM glucose. Samples were then incubated at 37uC in KRB-HEPES buffer (pH 7.4) with 1% bovine serum albumin and 2.8 mM glucose for 1 hour before insulin stimulation with 2.8 or 20 mM glucose. At the end of the 30 min stimulation, the supernatant was assayed for secreted insulin (by MSD), and cells/islets were assayed for total insulin and total protein.

RNA isolation and gene expression measurement
RNA was extracted from pancreas, islets or cultured cells using RNeasy Mini Kit (Qiagen) and quantified by RiboGreen RNA Assay Kit (Invitrogen). Reverse transcription was performed using qScript cDNA supermix (Quanta). Quantitative mRNA measurement was carried out by using qPCR core kit for SYBR Green I (Quanta) with the StepOne Plus detection system (Applied Biosystems). Levels of Xbp-s (spliced form) were normalized to Xbp-t (total) levels. Other gene expression levels were normalized to Gapdh and Actin levels of the same sample. Mouse primer sequences were listed as follows: Actin

Measurement of b-cell proliferation and volume
To measure b-cell proliferation, all experimental mice received BrdU at a concentration of 0.8 mg/ml in drinking water. Mice # P17 days were treated with BrdU water for 3 days, and mice $ 30 days were treated for 7 days. After BrdU administration, whole pancreata were harvested for immunohistochemistry following the standard procedure as previously described [30]. Guinea pig Antiinsulin (1:500, Abcam) and mouse anti-BrdU (1:50, DAKO) were applied overnight at 4uC. Anti-guinea pig and anti-mouse secondary antibodies conjugated with Alexa Fluor488 and 555 dye (Molecular Probes) were used (1:400 dilution) to visualize the labeled cells. Anti-fade reagent with Dapi (Life technologies) was used to mount slides and label the nucleus. Fluorescence images were captured with a Nikon Eclipse E1000 and Image-Pro Plus (Phase 3 Imaging Systems, GE Healthcare, Inc.). To calculate bcell daily proliferation, ratios of BrdU-positive to total b-cells were counted from a total of 4 tissue sections per mouse and further divided by total days of BrdU treatment. The same images were used for b-cell volume estimates. Insulin-positive cells were traced by using imageJ software, and single cell volume was estimated by dividing the total area of b-cells by cell number.

Estimation of b-cell number
Total b-cell number was estimated by a method we used previously [14,31]. The total amount of Glut2 mRNA or Insulin II mRNA in whole pancreata was first determined and then divided by the estimated amount of Glut2 or Insulin II mRNA, respectively, per b-cell. Since these two genes are exclusively expressed in b-cells of the endocrine pancreatic islets, the ratio of their total pancreatic level to single cell level can be used to estimate total b-cell number [14,31]. Similarly, b-cell number was estimated by dividing insulin protein level in whole pancreata by insulin content per b-cell.

Statistical Analysis
All numerical data were represented as mean 6 SE. Statistical significance was determined using Student's t testing.

Perk heterozygous mice exhibit increased levels of circulating insulin and decreased glucose
Mice and humans that are completely deficient for PERK exhibit severe hyperglycemia (.400 mg/dl) even in the fasted state [12,14,15,19,26]. Therefore we expected that half dosage of PERK associated with Perk heterozygosity would either be recessive with no effect on glucose homeostasis or would be semi-dominant with elevated blood glucose. Surprisingly we found that adult Perk +/2 mice exhibited significantly lower random-fed serum glucose levels than mice homozygous for the normal (Perk +/+ ) wild-type allele ( Fig 1A). The mouse Perk KO allele used in this study was generated by us [15] and utilized in several previous studies [13][14][15]20,25,27,30,[32][33][34][35][36][37]. Our Perk KO allele was made by Cremediated deletion of exons 5-7 that encoded part of the luminal domain, the transmembrane domain, and part of the catalytic domain of PERK yielding a loss-of-function null allele. Consistent with this, Perk +/2 mice exhibited the expected reduction in Perk mRNA and protein (ca. 60% normal, P,0.001) (Fig 1B-C) levels compared to wild-type Perk +/+ mice. Moreover, reduction in PERK activity was observed in Perk +/2 b-cells as phosphorylation of eIF2a, a downstream substrate of PERK, was significantly reduced by 30% (P,0.01) ( Fig 1D). Although circulating glucose levels and PERK activity are reduced, the islet composition and b-cell morphology in the pancreata of Perk +/2 is indistinguishable from Perk +/+ and these observations are in contrast to radical changes seen Perk KO mice, which include reduced b-cell number (Fig 1E). Perk +/2 b-cells also show normal enrichment of proinsulin in the Golgi (not shown) in contrast to ER retention seen Perk KO b-cells [38].
To determine when in development Perk +/2 mice first display mild hypoglycemia, we analyzed a large cohort of mice at various postnatal stages (Fig 2A). The genetic background of these mice was congenic C57Bl/6. Perk +/2 and Perk +/+ littermates were used throughout this study to reduce inter-litter variation. Blood glucose levels of neonatal (postnatal day 5) Perk +/2 mice were indistinguishable compared to Perk +/+ , but as mice approached the weaning stage of development (postnatal day 17) a substantial reduction in blood glucose levels was seen in Perk +/2 mice. Curiously, the difference in blood glucose between Perk +/2 and Perk +/+ mice was reduced with age. By the time the mice reached postnatal 75 days, the difference in blood glucose was no longer statistically significant. However, male Perk +/2 mice showed a persistent trend towards reduced blood glucose as seen in 9-12 month old mice in three genetic backgrounds ( Fig 2B) including a highly statistically significant difference for Perk +/2 in a 129 SvEvTac genetic background (Fig 2B, P,0.01). Serum insulin levels were inversely proportional to serum glucose levels with Perk +/2 mice (C57Bl/6 genetic background) showing significantly higher levels than Perk +/+ at P17 (P,0.05, Fig 2C) and a nonsignificant trend towards higher insulin levels in older mice.
To determine if Perk dosage would impact serum glucose and insulin under glucose stimulation, a glucose tolerance test (GTT) was performed in juvenile (P17) and adult mice (P50). Perk +/2 mice trended towards being more glucose tolerant (P = 0.1) than WT in P17 mice ( Fig 2D) and were significantly more glucose tolerant in P50 mice (P,0.05, Fig 2F). Serum insulin exhibited a strong trend towards being higher in Perk +/2 mice injected with glucose ( Fig 2E) but was not quite statistically significant.
Insulin content and b-cell are modulated by Perk gene dosage during postnatal development To determine whether insulin synthesis and storage contribute to PERK-dependent regulation of glucose and insulin homeostasis, whole pancreatic insulin content was measured. Compared to Perk +/+ , Perk +/2 mice exhibited 42.8% and 92.8% increase of pancreatic insulin at P17 and P50 (P,0.05 at both ages) (Fig 3A), respectively. By contrast, there was no Perk-dependent difference at other developmental time points, suggesting a developmental complexity underlying the Perk genotype differences in glucose homeostasis. These observed differences in total pancreatic insulin could arise from either differences in insulin content per b-cell or differences in total b-cell number. We estimated cell volume and insulin concentration per b-cell, both of which determine the insulin content per b-cell. b-cell volume did not differ between genotypes at any developmental stage (Fig 3B), suggesting that Perk genotypic difference in insulin content is due to a difference in bcell insulin concentration. We used the concentration of insulin per islet cells as an estimate of the insulin concentration per b-cell, given the fact that b-cells comprise most of the islet mass and that cell-type composition of islets is not different between genotypes (data not shown). At P17, b-cell insulin concentration was significantly higher in Perk +/2 compared to Perk +/+ (P,0.05, Fig 3C), suggesting that the elevated whole pancreatic insulin content at P17 was due to increased cellular insulin production and/or storage. The per-b-cell insulin concentration was equivalent in Perk genotypes at P5, P30, and P50. Interestingly, by contrast to juvenile mice, 6-month old (P180) Perk +/2 mice showed significantly reduced insulin concentration per b-cell revealing an unexpected developmental complexity. We also estimated pancreatic b-cell number by dividing the total amount of insulin in whole pancreata by the estimated insulin content per b-cell. We found that Perk +/2 mice initially had fewer b-cells during neonatal development, but this trend was reversed in mature adult mice ( Fig 3D). In summary, neonatal Perk +/2 mice exhibit higher insulin content per b-cell, enhanced insulin synthesis but fewer b-cells, whereas in mature adult Perk +/2 mice insulin concentration is reduced, while b-cell number is increased.
b-cell number is increased in Perk heterozygotes due to elevated b-cell proliferation Unlike P17 mice, Perk +/2 mice at P50 did not show increased expression of insulin mRNA level (Fig 4B) or protein level per b-cell ( Fig 3C). However, P50 Perk +/2 mice had substantially higher bcell number (Fig 3D). To confirm this observation, b-cell number was estimated using the expression of mRNA of two genes, insulin II and Glut2, after previously published methods [38,39]. Since both genes are exclusively expressed in b-cells, their mRNA levels in whole pancreata are directly proportional to b-cell number [38,39]. Perk +/2 mice at P50 had higher total insulin (P,0.05, Fig 4A) and total Glut2 (p = 0.08) mRNA in the total pancreas compared to wild-type mice whereas expression levels of these two genes in islets were not different between genotypes (Fig 4A), reflecting a 56%-69% increase in total b-cells in Perk +/2 (Fig 4A) with equivalent level of expression of insulin II and Glut2 per b-cell.
To further investigate the reason for increased b-cell number in P50 Perk +/2 mice, b-cell proliferation was determined by BrdU incorporation. b-cell proliferation was found to be significantly  increased in P50 Perk +/2 mice compared to WT controls (Fig 4B). We also examined b-cell proliferation at four other developmental time points and found that Perk +/2 exhibited elevated proliferation at P30 and P50 but not earlier or later time points (Fig 4B), indicating that enhanced proliferation was transient and corresponded to the time period when b-cell number was increased in Perk +/2 mice. In addition, b-cell death was estimated using TUNEL assay and found to be negligible and not different between Perk genotypes (data now shown).
Insulin transcription and proinsulin synthesis were upregulated in Perk+/2 mice at postnatal day 17 Despite a lower number of b-cells, P17 Perk +/2 mice exhibited higher pancreatic insulin due to a significant increase in insulin content per b-cell (Fig 3A-D). To probe the mechanism underlying the increased b-cell insulin content in Perk +/2 P17 mice, Insulin mRNA was measured in Perk +/2 islets and found to be 21% higher (P,0.05) than WT. To determine if increased Insulin gene transcription was responsible for the increased steadystate levels of Insulin mRNA, Insulin pre-mRNA, which has a much shorter half-life and is less abundant than the mature mRNA, was measured using primers detecting the Insulin intron after methods of Evans-Molina and coworkers [40]. Insulin pre-mRNA was elevated 52% in Perk +/2 b-cells (P,0.05), whereas Glucagon mRNA was not impacted by modulation of Perk (Fig 5A), suggesting that Perk-dependent difference in Insulin gene transcription contributed to the difference in insulin content. Unlike stage P17 mice, no change of mature mRNA or pre-mRNA of insulin was seen in Perk +/2 b-cells at other developmental time points (Fig 5B).

ER chaperones are differentially regulated by Perk gene dosage
To determine if the expression of other genes associated with insulin biosynthesis exhibited Perk genotypic differences in mice at P17, mRNA levels were determined in isolated islets for MafA, Pdx1, Hrd1, ERp57, BiP, and ERp72. MafA mRNA was increased by 25% (p = 0.06) in Perk +/2 whereas Pdx1 was not changed (Fig 7A). The expression of the mRNAs encoding the ER chaperones HRD1, BIP, and ERp72 levels were significantly elevated in Perk +/2 b-cells, while ERp57 mRNA was reduced (Fig 7A). The expression of the same genes was measured in isolated islets of mice at P30 (Fig 7B) and P50 (Fig 7C) but none showed a genotypic difference. The protein levels of BIP, ERp72, ERp57 and another protein disulfide isomerase family protein PDI  were assessed in pancreatic islets of P17 mice. ERp72 was elevated Perk +/2 islets (Fig 7D), whereas PDI was decreased and BIP and ERp57 were not different from levels seen in Perk +/+ islets. In addition, we also measured mRNA level of Chop, Atf4 and Xbp-1 splicing in P17 mouse islets, which are sensitive indicators of ER stress. None of the ER stress markers showed Perk genotypic differences (Fig 7A), suggesting that regulation of b-cell functions by Perk dosage was not mediated through ER stress pathway.
Perk gene dosage specifically in the pancreatic b-cells regulates glucose homeostasis Although our analysis of b-cell functions suggests that the Perk genotypic differences in glucose homeostasis are due to differences in expression levels of PERK in b-cells, other organs that are known to regulate glucose homeostasis, including the liver, may also participate in this regulation. To pinpoint the responsible organ/cell type, we generated mouse strains in which Perk gene dosage was altered in specific organs and/or cell types. Examination of liver specific Perk KO (liPKO) mice revealed no differences in random fed glucose levels (Fig 8A). By contrast, we previously reported that pancreatic specific Perk KO (pcPKO) rapidly developed severe hyperglycemia similar to global Perk KO mice [14]. In addition, we now report that pcPKO heterozygotes exhibit 21% (P,0.01) lower random fed glucose levels than corresponding wild-type control in mice 3-5 weeks old (Fig 8B), suggesting that reducing Perk gene dosage in half specifically in the pancreas recapitulates the reduced serum blood glucose seen in the Perk heterozygous mice. Consistent with these observations, mice expressing an extra copy of Perk specifically in b-cells with an otherwise wild-type background (Perk +/+ ;bPerk) exhibited significantly elevated serum glucose (P,0.05, Fig 8C) and reduced serum insulin (P,0.001, Fig 8D). Therefore, the effect of Perk gene dosage on insulin and glucose homeostasis is likely to be b,cell specific.
We sought to reduce Perk expression in cultured b-cells to confirm the importance of Perk gene dosage in b-cells and the difference in insulin synthesis and secretion we observed in Perk +/2 mice. To accomplish this we modulated Perk mRNA in INS1 832/ 13 b-cells through regulating the expression of a stably integrated shPerk transgene under the control of doxycycline (denoted as INS1 832/13 shPerk cells). After 24-hour administration of various concentration of doxycycline ranging from 0 to 2 mg/ml, Perk mRNA level was modulated within a range of 39.7%-100% of normal (Fig 8E). Maximum knockdown of Perk mRNA was achieved by using 2 mg/ml doxycycline. After 24-hour treatment of 2 mg/ml doxycycline, cells exhibited impaired GSIS and significantly elevated ERp72 expression (Fig 8F and 8G), which were consistent with previous observations in mice or culture cells with total ablation of PERK by other means [14,36]. By examining the dose-response curve, we found that the application of 0.002 mg/ml doxycycline for 24 hours provided a 40% reduction in Perk mRNA (P,0.001, Fig 8E) that mimicked the levels observed in Perk +/2 mice (Fig 1B). Using this strategy, we found that both glucose stimulated insulin secretion and ERp72 gene expression were significantly elevated in shPerk cells treated with 0.002 mg/ml doxycycline for 24 hours (Fig 8F and 8G), which was consistent with our observations in Perk +/2 b-cells.

Discussion
A complete deficiency of PERK results in the severest form of insulin-dependent diabetes [12,15,19], and therefore we expected that Perk heterozygosity would either be recessive with no effect on glucose homeostasis or would be semi-dominant with reduced insulin and elevated blood glucose. Unexpectedly, we found that Perk heterozygous mice exhibit an over-dominant phenotype in early postnatal development characterized by elevated insulin and correspondingly reduced blood glucose levels and increased glucose clearance. By using tissue or cell specific Perk KO mice and b-cell targeted Perk transgene we previously demonstrated that the insulin insufficiency and complete loss of glucose homeostasis was caused by the absence of PERK in the b-cells. Using one of these strains we generated pancreatic-specific Perk heterozygotes and found that pcPerk +/2 mice had reduced blood glucose similar to Perk +/2 . Once again, this was opposite to what we expected based on the diabetic phenotype of pcPerk +/2 mice. Previously we showed that wild-type Perk transgene exclusively targeted to be expressed in b-cells could reverse the diabetes of the Perk KO mouse [14,27]. However, when this transgene is present in an otherwise wild-type (Perk +/+ ) background it results in the reduction of serum insulin and the elevation of blood glucose. Thus, circulating insulin and blood glucose levels are negatively and positively correlated, respectively, with Perk gene dosage in the pancreatic b-cells. The effect of Perk gene dosage on glucose homeostasis is amplified in combination with the dominant Akita insulin mutant, which progressively develops diabetes postnatally.
Lower Perk gene dosage slows the progression of diabetes in the Akita mouse whereas overexpression of Perk specifically in b-cells hasten it [20]. The stark exception to this rule is when Perk gene dosage is equal to zero.
The expression of PERK in the liver has been suggested to play an important role in glucose homeostasis in the first few days of life when gluconeogenesis plays a crucial role in providing glucose to the neonates [15,23]. However, we found that glucose homeostasis was unaffected by genetically deleting the Perk gene in the adult liver. Consequently we assert that the effect of Perk gene dosage on insulin and glucose homeostasis is unlikely to be dependent upon liver functions and is only dependent upon the relative expression of Perk in the insulin-secreting b-cells as also supported by direct manipulation of Perk gene dosage therein.
Comparison of serum insulin and glucose levels throughout postnatal development shows a simple inverse relationship (Fig 9A). Given that we found no evidence for differences in peripheral insulin sensitivity before six-months of age, we conclude that Perk genotypic differences in blood glucose are directly determined by the amount of insulin secreted by the pancreatic b-cells. However the underlying reasons for elevated insulin secretion in Perk heterozygous mice change during postnatal development. Initially, as seen at postnatal day 17, total pancreatic insulin is elevated despite a reduced b-cell number indicating that each b-cell has substantially more stored insulin (Fig 9B). Later b-cell proliferation is accelerated in Perk heterozygotes. Although this acceleration is modest and transient, the compounding effect of increased proliferation over three weeks leads to a significant accrual of bcell number in Perk +/2 mice. The relative large b-cell number in Perk heterozygotes is maintained thereafter. However, as b-cell number increases insulin content per b-cell drops resulting in no genotypic difference in total pancreatic insulin in mice beyond 7 weeks. One constant observed across all ages is an elevation in the amount of insulin secreted per b-cell in Perk heterozygotes.
Recently we showed that PERK acutely regulates calcium dynamics and insulin secretion in human and rodent b-cells [25] independently of the eIF2a pathway. These experiments were performed using a PERK inhibitor that allowed us to determine the immediate effect of PERK on the crucial steps in calcium mobilization and insulin secretion. We concluded that PERK acts to regulate calcium dynamics and insulin secretion independently of its well-known role in phosphorylating the translation initiation factor eIF2a [25]. PERK has also been shown to regulate proinsulin quality control and trafficking in the endoplasmic reticulum [20], which is dependent on the phosphorylation of eIF2a by PERK. In the absence of PERK, proinsulin and ER client proteins eventually accumulate to extremely high levels in the ER and the ER ceases to function. The function of PERK in regulating ER quality control and trafficking is likely to be associated with its phosphorylation of eIF2a, as mutants of the regulatory phosphorylation site of eIF2a results in the same cellular phenotypes in b-cells [23]. These findings support the hypothesis that PERK has multiple functions in the pancreatic bcells including immediate regulation of calcium dynamics and insulin secretion and long term regulation of the ER chaperones that orchestrate quality control, protein folding, and anterograde trafficking to distal compartments of the secretory pathway. In addition, PERK regulates b-cell proliferation as first demonstrated in Perk KO mice and now shown in Perk heterozygous mice [14]. It is unclear, whether the transient increase in b-cell proliferation in Perk heterozygotes is a primary function of PERK or is in response to changes in other b-cell functions.
The complex behavior of b-cells in Perk heterozygotes over the first few months of postnatal development can be subdivided into direct effects of PERK expression differences and adaptive responses to primary effects. Insulin secretion is likely to be a direct effect of PERK expression changes because it has been shown to be acutely regulated by PERK [25]. The relationship between insulin secretion and PERK expression exhibits a biphasic, inverted U-shaped dose response with half-dosage (Perk +/2 ) defining the maximum. This unexpected biphasic relationship between PERK expression and insulin secretion can account for the unexpected lower blood glucose levels seen in Perk heterozygotes. A similar inverted dose response relationship was noted between PERK expression and regulation of intermediary metabolism genes in a cell culture system [41]. More mysterious is the transient elevation of b-cell proliferation in Perk heterozygotes, which is unlikely to be due to a normal compensatory response in response to hyperglycemia and an increase demand for insulin similar to that seen in the progression to type 2 diabetes [42]. Previous studies suggested that intracellular Ca 2+ positively regulates b-cell proliferation [43][44][45]. Given the fact that elevated insulin secretion continues to be observed in P30 and P50 Perk +/2 mice, the increased b-cell proliferation in P30 and P50 Perk +/2 mice might be due PERK-dependent regulation of Ca 2+ signaling and insulin secretion. Regardless of the reason, the enhanced bcell proliferation results in an increase in b-cell mass in Perk +/2 mice, which is maintained indefinitely in the mature adult. The decrease in stored insulin that is seen in older adult mice (. 2 months old) is strongly correlated with the increased b-cell mass and is therefore likely to be a compensatory response to avoid hyperinsulimia. Among the five postnatal stages studied, P17 exhibited the largest array of differences including insulin gene transcription, proinsulin synthesis, insulin content, insulin secretion, hypoglycemia, b-cell mass, and changes in the expression of ER chaperones. It is likely that subsequent changes represent adaptive response in order to maintain glucose homeostasis.
Harding and Ron [12] independently generated a Perk KO mouse, which exhibited the same phenotype as our Perk KO strain including diabetes. Harding and Ron reported that older Perk Genotypic difference of serum glucose and insulin throughout development. Figure 9A was generated based on the data in Fig 2A and Fig 2C. Serum glucose and insulin are inversely correlated during postnatal development of Perk +/2 mice. B. Genotypic difference of total pancreatic insulin content (based on data in Fig 3A), insulin content per b-cell (Fig 3C), b-cell number (Fig 3D), and b-cell proliferation rate (Fig 7B). Initially b-cell number is relatively low in Perk +/2 , but b-cell and total pancreatic insulin are high. In response to low b-cell number, b-cell proliferation is accelerated between postnatal 17-50 resulting in increased b-cell number in Perk +/2 . However, as b-cell number rises, insulin content per b-cell drops resulting in a balance between cell number and insulin concentration in each cell and a return to equivalent levels of total pancreatic insulin content. doi:10.1371/journal.pone.0099684.g009 heterozygous mice (6 months old) exhibited mild glucose intolerance in three different genetic backgrounds and this defect was stable from 6 months to 1 year old [12]. Although our Perk heterozygous mice show the opposite phenotype in terms of glucose tolerance in younger mice, they exhibit mild glucose intolerance at 6 months of age (data not shown). Moreover, the highly significant difference in random fed glucose levels between Perk genotypes observed in younger mice gradually diminishes with age and becomes non-significant. Taken together we speculate that the higher level of circulating insulin that we observed in Perk heterozygous mice during early postnatal development eventually lead to insulin resistance in older mice analogous to the progression of type 2 diabetes from compensation to decompensation [42].
In conclusion, the complex and dynamic regulation of b-cell functions revealed by investigating Perk heterozygotes strongly argues that PERK has important physiological and postnatal developmental functions in b-cells. These functions are not likely to entail the well-known function of PERK to regulate the ER stress response in cultured cells as markers for ER stress were not differentially expressed.