Critical reduction in beta-cell mass results in two distinct outcomes over time. Adaptation with impaired glucose tolerance or decompensated diabetes.

We have proposed that hyperglycemia-induced dedifferentiation of beta-cells is a critical factor for the loss of insulin secretory function in diabetes. Here we examined the effects of the duration of hyperglycemia on gene expression in islets of partially pancreatectomized (Px) rats. Islets were isolated, and mRNA was extracted from rats 4 and 14 weeks after Px or sham Px surgery. Px rats developed different degrees of hyperglycemia; low hyperglycemia was assigned to Px rats with fed blood glucose levels less than 150 mg/dl, and high hyperglycemia was assigned above 150 mg/dl. beta-Cell hypertrophy was present at both 4 and 14 weeks. At the same time points, high hyperglycemia rats showed a global alteration in gene expression with decreased mRNA for insulin, IAPP, islet-associated transcription factors (pancreatic and duodenal homeobox-1, BETA2/NeuroD, Nkx6.1, and hepatocyte nuclear factor 1 alpha), beta-cell metabolic enzymes (glucose transporter 2, glucokinase, mitochondrial glycerol phosphate dehydrogenase, and pyruvate carboxylase), and ion channels/pumps (Kir6.2, VDCC beta, and sarcoplasmic reticulum Ca(2+)-ATPase 3). Conversely, genes normally suppressed in beta-cells, such as lactate dehydrogenase-A, hexokinase I, glucose-6-phosphatase, stress genes (heme oxygenase-1, A20, and Fas), and the transcription factor c-Myc, were markedly increased. In contrast, gene expression in low hyperglycemia rats was only minimally changed at 4 weeks but significantly changed at 14 weeks, indicating that even low levels of hyperglycemia induce beta-cell dedifferentiation over time. In addition, whereas 2 weeks of correction of hyperglycemia completely reverses the changes in gene expression of Px rats at 4 weeks, the changes at 14 weeks were only partially reversed, indicating that the phenotype becomes resistant to reversal in the long term. In conclusion, chronic hyperglycemia induces a progressive loss of beta-cell phenotype with decreased expression of beta-cell-associated genes and increased expression of normally suppressed genes, these changes being present with even minimal levels of hyperglycemia. Thus, both the severity and duration of hyperglycemia appear to contribute to the deterioration of the beta-cell phenotype found in diabetes.

Pancreatic ␤-cells maintain specialized pathways of metabolism that efficiently couple the secretion of insulin to circulating glucose levels (1,2). With the increased demand of insulin resistance and obesity, an adaptation in ␤-cell mass and secretion can keep glucose levels within a narrow range (3)(4)(5). The failure of ␤-cells to adequately adapt to increased demand is fundamental to the pathogenesis of all forms of diabetes. We have hypothesized that abnormal ␤-cell function in diabetes is due to the loss of the unique expression pattern of genes that optimize glucose-induced insulin secretion (GIIS) 1 and insulin synthesis (5).
The development of the endocrine pancreas and the maintenance of ␤-cell differentiation is regulated by a network of transcription factors, which includes pancreatic and duodenal homeobox-1 (PDX-1), BETA2/NeuroD, Nkx6.1, and hepatocyte nuclear factors (HNFs). These factors regulate transcription of the insulin gene and genes involved in ␤-cell glucose sensing such as GLUT2 and glucokinase (6 -12). Optimal ␤-cell function is dependent on expression of these genes and the suppression of other genes, including lactate dehydrogenase-A (LDH-A), hexokinase I, and enzymes required for gluconeogenesis, that can be predicted to interfere with optimal secretion.
In this study, we have extended our previous findings examining the influence of the diabetic milieu on ␤-cell differentiation in the rat partial pancreatectomy (Px) model of diabetes (13)(14)(15). Active regeneration is seen during the first 7-10 days following surgery, but by 14 days the morphology of the pancreas has stabilized, appearing similar to what is seen at much later time points. We have found that at the 4-week time point after surgery there is a loss of ␤-cell differentiation associated with insulin secretory defects and ␤-cell hypertrophy (13)(14)(15)(16)(17). Questions have been raised as to whether this 4-week time point provides a representative view of islet adaptation to the diabetic milieu or whether the findings resulted from some regeneration artifact. For these reasons, islet phenotype and morphology were examined at both 4 and 14 weeks following partial pancreatectomy.

EXPERIMENTAL PROCEDURES
Animals-Male Sprague-Dawley rats (Taconic Farms, Germantown, NY), weighing 90 -100 g were submitted to 85-95% Px or sham Px surgery as previously described (13,18). For partial pancreatectomy, tissue removal was performed by gentle abrasion with cotton applicators leaving the pancreas within 1-2 mm of the common pancreatic bile duct and extending from the duct to the first part of the duodenum. The proportion of the gastric lob removed was varied to generate 85-95% Px rats that develop different degrees of hyperglycemia (13). For sham surgery, the pancreatic tissue was only lightly rubbed instead of being removed. Animals were weighed, and blood was obtained in heparinized microcapillary tubes from snipped tails of fed rats (9 -10 a.m.) weekly. Whole-blood glucose levels were measured with a portable Medisense Precision QID glucometer (Abbott Laboratories, Bedford, MA). Rats were classified according to their averaged blood glucose levels from 3 weeks after surgery; low hyperglycemia (LPx) was assigned to Px rats with blood glucose levels less than 150 mg/dl, and high hyperglycemia (HPx) was assigned above 150 mg/dl. To assess the effects of the duration of hyperglycemia on islet gene expression, rats were anesthetized at 4 and 14 weeks after surgery, and their islets were isolated with collagenase digestion of the pancreatic remnant or sham pancreas. Islets were further separated with a Histopaque density gradient (Histopaque-1077; Sigma) and hand-picked under a stereomicroscope. Islets of similar size were used for extraction of RNA. In several cases, it was necessary to pool islets from two Px rats with similar glycemic levels to obtain an islet yield that was sufficient for RNA extraction. Animals were kept under conventional conditions with free access to water and standard pelleted food. All animal procedures were approved by the Joslin Diabetes Center Animal Care Committee.
RNA Extraction and Synthesis of cDNA-Total RNA was extracted from islets using Ultraspec RNA isolation reagent according to the manufacturer's suggested protocols (Biotecx Laboratories, Houston, TX) and quantified by spectrophotometry. RNA (500 ng) was reversetranscribed into cDNA in a final reaction solution of 25 l containing the following: 1ϫ Superscript first strand buffer (50 mM Tris-HCl, 75 mM KCl, and 3 mM MgCl 2 ) (Invitrogen), 40 units of Rnasin (Promega, Madison, WI), 10 mM dithiothreitol, 1 mM dNTP, 50 ng of random hexamers, and 200 units of Superscript II Rnase H Ϫ reverse transcriptase (Invitrogen). Reverse transcription reactions were incubated for 10 min at 25°C, 60 min at 42°C, and 10 min at 95°C. Resultant cDNA products were diluted with 50 l of H 2 O to a concentration corresponding to 10 ng of starting RNA per 1.5 l.
Semiquantitative Radioactive Multiplex Polymerase Chain Reaction-PCRs were carried out in a volume of 25 l consisting of 10 ng of cDNA, 1ϫ GeneAmp PCR Gold buffer (Applied Biosystems, Foster City, CA), 1-2 mM MgCl 2 , 80-160 M dNTP, 80 -600 nmol of oligonucleotide primers (Sigma), 1.25 Ci of [␣-32 P]dCTP (3000 Ci/mmol; PerkinElmer Life Sciences), and 2.5 units of AmpliTaq Gold DNA Polymerase (Applied Biosystems, Foster City, CA). Table I shows specific concentrations of MgCl 2 , dNTP, and oligonucleotide primers, along with multiplex PCR conditions for each gene tested. Reactions were performed in a 9700 Thermocycler (Applied Biosystems, Foster City, CA) in which samples underwent a 10-min initial denaturing step, followed by the number of cycles indicated (Table I), for durations of 1 min at 94°C, 1 min at the annealing temperature indicated in Table I, and 1 min at 72°C. The final extension step was 10 min at 72°C. Amplimers were resolved by 6% polyacrylamide gel electrophoresis in 1ϫ Tris borate EDTA buffer. The amount of [␣-32 P]dCTP incorporated into amplimers was measured with a Storm 840 PhosphorImager and quantified with ImageQuant software (Amersham Biosciences). The average intensity of each product was expressed relative to the internal control gene (ratio of specific product/control gene). These ratios were then used to calculate the percentage of sham expression for each Px animal in the same reverse transcription-PCR. We have previously verified (13)(14)(15)) that the multiplex PCR products for each set of primers are linearly amplified. Control experiments were performed to adjust the PCR conditions such that the number of cycles used was in the exponential phase of amplification for all products and that each PCR product in a multiplex reaction increased linearly with the amount of starting material.
Plasma Insulin and Lipid Determination-Plasma insulin was measured by radioimmunoassay (Linco Research, St. Charles, MO). Plasma lipids were measured from samples collected in ETDA/diethyl-paranitrophenyl phosphate (paraoxan)-coated tubes to avoid activation of lipoprotein lipase by heparin (19). Plasma nonesterified fatty acids (NEFA) were measured by a colorimetric method (Wako Chemicals, Neuss, Germany). Plasma triglyceride were measured with a triglyceride assay kit (GPO Trinder; Sigma) using glycerol as a standard.
Phlorizin Treatment of Px Rats-To reverse hyperglycemia, Px rats were treated with phlorizin for the final 2 weeks of the 14-week study period. Phlorizin was dissolved in 1,2-propanediol and injected intraperitoneally twice a day (9 a.m. and 9 p.m.) at a dose of 0.8 g/kg/day. Sham animals received similar amounts of 1,2-propanediol as phlorizin-treated Px rats.
Cell Size-␤-Cell size was measured in sections of pancreas from sham and Px rats stained with glucagon (rabbit anti-bovine glucagon; gift of M. Appel, Worcester, MA) and counterstained with hematoxylin as previously described (17). The number of islet nuclei were counted (nuclei of cells staining for glucagon were excluded), and the area of these cells was quantitated using image analysis software (IP Lab Spectrum). Average cross-sectional ␤-cell area was determined as the area of the cells divided by the number of nuclei. An average of 725 Ϯ 96 cells/animal were measured.
Glucose Clamp-To assess the influence of short term hyperglycemia on islet gene expression in vivo, catheterized rats (ϳ250 g) were infused for 4 days with glucose (500 g/liter hydrated glucose, McGaw, Irvine, CA) or saline (4.5 g/liter) as described (20,21). The glucose infusion rate was regularly adjusted to maintain the blood glucose level at ϳ200 mg/dl, with saline infusion adjusted accordingly. At the end of the infusion period, islets were isolated, and RNA was extracted for reverse transcription-PCR analysis. Experiments were performed on islet extracts from rats used in our previous study (21).
Statistical Analysis-All results are presented as means Ϯ S.E. Statistical analyses were performed using unpaired Student's t test or one-way analysis of variance.

Changes in Fed Blood Glucose Levels and Body Weight after
Px-Time course changes in blood glucose levels after Px are shown in Fig. 1. As previously described (13)(14)(15), slight variation in the proportion of pancreas removed (ϳ85-95%) resulted in rats with different degrees of hyperglycemia after 4 weeks; blood glucose levels ranged from 114 to 370 mg/dl. However, after 14 weeks, the range of blood glucose levels in Px rats had clustered into two distinct groups with either low or high levels of hyperglycemia (Fig. 1). Px rats were classified into groups according to their averaged fed blood glucose levels; LPx below 150 mg/dl and HPx above 150 mg/dl. The time course changes in body weight and fed blood glucose in LPx, HPx, and sham-Px rats are illustrated in Fig. 2. As previously described (13,18), weight gain in Px rats was slightly decreased during the first few days after surgery, resulting in significantly lower body weights at 1 week postsurgery. Thereafter, Px rats gained weight at the same rate as sham Px rats. After 8 weeks, LPx rats had body weights not significantly different from agematched sham rats, whereas HPx rats had lower body weights throughout the study period. Averaged post-3-week fed blood glucose levels ranged from 97 to 111 mg/dl in sham rats, from 108 to 141 mg/dl in LPx rats, and from 220 to 365 mg/dl in HPx rats. The blood glucose levels of HPx rats were significantly increased by 1 week after surgery and remained significantly higher compared with sham and LPx rats throughout the study (p Ͻ 0.001). Blood glucose levels of LPx rats were not different from sham at 1 week after surgery; however, the averaged post-3-week blood glucose levels in LPx rats were significantly higher than sham (p Ͻ 0.01).
Changes in ␤-Cell Phenotype-The expression of islet-associated transcription factors, islet hormones, specialized ␤-cell metabolic enzymes, and ion channels/pumps along with other normally suppressed metabolic enzymes and stress genes were measured in islets from sham, LPx, and HPx rats sacrificed 4 and 14 weeks after surgery.
Changes in Levels of Islet-associated Transcription Factor mRNA-After normalization of the specific gene to an internal control gene (TBP, ␣-tubulin, cyclophilin, or 18 S rRNA), mRNA levels in LPx and HPx islets were quantitated as a percentage of sham (Table II). The expression levels of several transcription factors (PDX-1; also known as IDX-1, IPF-1, and STF-1), ␤-cell E-box trans-activator 2 (BETA2/NeuroD),

␤-Cell Abnormalities Worsen with Duration of Hyperglycemia
Nkx6.1, and hepatocyte nuclear factor 1␣ (HNF1␣)) important for pancreas and islet development and the maintenance of ␤-cell differentiation were assessed in Px islets. After 4 weeks, the mRNA levels of PDX-1, BETA2/NeuroD, Nkx6.1, and HNF1␣ were unaltered in islets from LPx rats, whereas in HPx rats they were reduced by 50 -60% compared with sham (Table  II). After 14 weeks, the mRNA levels for PDX-1, Nkx6.1, and HNF1␣ were significantly reduced in LPx rats (30 -40%) and were further reduced in HPx rats (50 -60%). Thus, expression levels in LPx rats were unaltered at 4 weeks but significantly reduced after 14 weeks. Therefore, the down-regulation of isletassociated transcription factors after Px showed an association with increasing blood glucose levels and the duration of hyperglycemia. On the other hand, c-Myc is a transcription factor minimally expressed in replicating cells, which is consistent with its low level expression in adult islets (13,22). After 4 weeks, c-Myc mRNA levels were unaltered in LPx rats but were increased in HPx rats (Table II). After 14 weeks, c-Myc mRNA levels were significantly increased in LPx rats and were increased further in HPx rats compared with sham control rats (p Ͻ 0.05 among groups). Therefore, the up-regulation of c-Myc was associated with increasing blood glucose levels and the duration of hyperglycemia.
mRNA Levels of Islet Hormones-Insulin and IAPP expression were down-regulated after Px with a similar dependence on the degree and duration of hyperglycemia. After 4 weeks, both insulin and IAPP mRNA levels were unaltered in LPx rats but were reduced in HPx rats (Table II). After 14 weeks, both genes were significantly reduced in LPx rats and were further reduced in HPx rats (insulin reduced by ϳ55%). In contrast, the mRNA levels for the ␣-cell hormone, glucagon, and the ␦-cell hormone, somatostatin, were unchanged in Px rats at both time points and thus showed no association with glycemia.
mRNA Levels of Metabolic Enzymes-Strikingly, the downregulation of several ␤-cell-associated metabolic enzymes (glu-cose transporter 2 (GLUT2), glucokinase, mitochondrial glycerol phosphate dehydrogenase, and pyruvate carboxylase) after Px showed a similar association with increasing blood glucose and the duration of hyperglycemia. After 4 weeks, mRNA levels were unaltered in LPx rats but were reduced by 40% in HPx rats (Table II). After 14 weeks, the ␤-cell-associated metabolic enzymes were, individually, not significantly changed in LPx rats. However, when the mRNA levels for this group of genes (GLUT2, glucokinase, mitochondrial glycerol phosphate dehydrogenase, and pyruvate carboxylase) were averaged for each islet RNA preparation, the averaged mRNA levels in LPx rats were significantly reduced compared with sham (25%, p Ͻ 0.05). They were further reduced (40 -50%) in HPx rats at 14 weeks. In contrast, several metabolic enzymes normally suppressed in ␤-cells (hexokinase I, LDH-A, and the gluconeogenic enzyme, glucose-6-phosphatase) were markedly increased in HPx rats at 4 and 14 weeks (Table II). In LPx rats, hexokinase I, LDH-A, and glucose-6-phosphatase were unaltered at 4 weeks. However, at 14 weeks, LDH-A mRNA levels were significantly increased, and hexokinase I and glucose-6-phosphatase tended to be increased, although not significantly.
mRNA Levels of Ion Channels/Pumps-The expression of several ion channels important for the stimulation of secretion were assessed in Px rats (Table II). Analyzed individually, the mRNA levels of the pore-forming subunit of the ATP-sensitive K ϩ channel Kir6.2, the ␤ subunit of voltage-dependent calcium channel, and sarcoplasmic reticulum Ca 2ϩ -ATPase 3 were not significantly reduced in LPx rats at 4 or 14 weeks. However, mRNA levels for this group of ion channels (Kir6.2, the ␤ subunit of voltage-dependent calcium channel, and sarcoplasmic reticulum Ca 2ϩ -ATPase 3) averaged for each experiment were significantly reduced in LPx rats at 14 weeks (25%, p Ͻ 0.05) but not at 4 weeks. In HPx rats, the mRNA levels of Kir6.2, the ␤ subunit of the voltage-dependent calcium channel, and sarcoplasmic reticulum Ca 2ϩ -ATPase 3 were each reduced to a similar extent at both 4 and 14 weeks.
mRNA Levels of Stress Genes-Stress gene mRNA levels in ␤-cells are considerably lower than in the liver and other tissues (23,24). mRNA levels of the antioxidant enzyme heme oxygenase-1 (HO-1) were unchanged in LPx rats at 4 weeks but were significantly increased at 14 weeks. Expression of both antioxidant genes, HO-1 and glutathione peroxidase were markedly increased in islets of HPx rats at 4 and 14 weeks. mRNA levels of the antiapoptotic gene, A20, were unaltered in LPx rats but were increased in HPx rats at 4 weeks. At 14 weeks, A20 expression was increased to a similar extent in LPx and HPx rats. Interestingly, expression of the proapoptosis cell surface protein, Fas, was similarly altered; it was unchanged in LPx but increased in HPx rats at 4 weeks and increased to similar levels in LPx and HPx rats at 14 weeks.
Changes in Plasma Insulin, NEFA, and Triglyceride Levels after Px-Changes in plasma insulin, NEFA, and triglyceride levels 4 and 14 weeks after surgery are shown in Fig. 3. At both time points, plasma insulin levels (Fig. 3A) tended to be decreased in Px rats but were only significantly reduced in HPx rats. Plasma NEFA (Fig. 3B) and triglyceride (Fig. 3C) levels were unchanged at 4 weeks regardless of the level of hyperglycemia. At 14 weeks, NEFA and triglyceride levels were unchanged in LPx rats but modestly increased in HPx rats.
Reversibility of Changes in Gene Expression after Px-We tested the reversibility of the changes in mRNA levels by using phlorizin, an inhibitor of glucose reabsorption in the kidney. Phlorizin blocks the Na ϩ /glucose co-transporter in the proximal tubules of the kidney, causing glucosuria and the normalization of circulating glucose levels. We previously showed (13-15) that expression levels of islet transcription factors,

␤-Cell Abnormalities Worsen with Duration of Hyperglycemia
hormones, metabolic enzymes, ion channels, and stress genes were completely normalized in 4-week Px rats after treatment with phlorizin. Here, we tested whether the altered islet gene expression in 14-week Px rats would be similarly normalized with 2-week phlorizin treatment. Blood glucose levels in Px rats were completely normalized during phlorizin treatment (Fig. 4). Expression of the transcription factors, PDX-1 and Nkx6.1, shown previously to be fully normalized with phlorizin treatment of 4-week Px rats (13), were only partially reversed toward normal at 14 weeks, remaining significantly reduced compared with sham levels (Fig. 5). Similarly, insulin and GLUT2 expression were partially restored after phlorizin treatment; GLUT2 mRNA levels were significantly decreased in phlorizin-treated Px rats compared with sham. The up-regulated expression of c-Myc, LDH-A, HO-1, and glutathione peroxidase in Px rats tended to be partially reversed after phlorizin treatment (Fig. 5). However, LDH-A, HO-1, and glutathione peroxidase mRNA levels were significantly increased in phlorizin-treated Px rats compared with sham. In all cases shown in Fig. 5, the vehicle had no effect on islet gene expression in sham-treated rats. Similar results were observed for the other genes altered in Table II (not shown).
Increased ␤-Cell Size in Px Rats-We previously found hypertrophy of ␤-cells in 4-week hyperglycemic Px rats (13,17). Here we confirmed these findings (Table III) and assessed the ␤-cell cross-sectional area 14 weeks after sham or Px (Table  III). Average ␤-cell cross-sectional area was increased by ϳ50% in 14-week Px rats. Thus, ␤-cell size was similarly increased at 4 and 14 weeks after Px. It is noteworthy that of the three Px rats of the 14-week group, all had increases of cell size, with two being very hyperglycemic but one having a fed blood glucose level of only 121 mg/dl.

Islet Gene Expression after Glucose Clamp-
The influence of short term hyperglycemia on islet gene expression was assessed using in vivo glucose clamps. Gene expression levels after 4-day glucose clamp (blood glucose maintained at ϳ200 mg/dl) relative to saline-infused control rats are shown in Table  IV. In contrast to the effects of long term hyperglycemia in Px rats (Table II), short term hyperglycemia induced by glucose clamp caused relatively little change in islet gene expression (Table IV). PDX-1 mRNA levels were unaltered after glucose clamp, and BETA2/NeuroD and Nkx6.1 mRNA were modestly reduced (Table IV). Insulin mRNA levels were unaltered after glucose clamp, whereas IAPP mRNA levels were increased consistent with previous findings in glucose-infused rats (20). Somatostatin mRNA levels were unaltered after glucose clamp, whereas glucagon mRNA was decreased, perhaps due the initial growth in ␤-cells, leading to a reduction in the proportion of ␣-cells per islet. Gene expression of representative ␤-cell metabolic enzymes, ion channels/pumps and stress genes were also tested in glucose-clamped rats; unchanged expression was found for GLUT2, Kir6.2, HO-1, glutathione peroxidase, and A20 (Table IV). We previously found increased c-Myc expression after glucose clamp (21). Note, as a caveat to the comparison between the effects of hyperglycemia in the short term after glucose clamp and the long term after Px, plasma insulin levels were higher during glucose clamp (21), whereas they tend to be lower after Px. DISCUSSION In this study, we have extended our previous findings examining the influence of the diabetic milieu on ␤-cell differentia-

␤-Cell Abnormalities Worsen with Duration of Hyperglycemia
tion in the rat 85-95% Px model of diabetes (13)(14)(15). Here we show that the changes in ␤-cell gene expression in Px rats are associated with both the severity and duration of hyperglyce-mia with the induction of several normally suppressed genes and decreased expression of genes that optimize ␤-cell function.
The global disruption of gene expression deteriorates over time, becoming resistant to reversal and evident even with minimal hyperglycemia, thus demonstrating the critical influence of hyperglycemia in the loss of ␤-cell differentiation. This association between chronic hyperglycemia and ␤-cell dedifferentiation provides a molecular mechanism for the ␤-cell dysfunction found in diabetes. It is also of interest that this critical reduction of ␤-cell mass results in two outcomes after 14 weeks. We noted a similar phenomenon with transplanted encapsulated mouse islets several years ago, finding that an identical number of islets either cured or did not cure diabetic mouse recipients with there being no middle ground (25). In the present Px study, a slight variation in the proportion of pancreas removed may have been responsible for the rats initially having a full range of blood glucose levels from low to middle and high values. However, over time, blood glucose levels clustered into two distinct groups. Some rats maintain nearly normal glucose levels that might be equated with the state of impaired glucose tolerance in humans. The other rats become highly hyperglycemic, with no rats maintaining glucose levels in the middle range (140 -250 mg/dl). Once decompensation occurs, glucose levels may rise to high levels, probably propelled by the combined forces of worsening glucotoxicity at the ␤-cell level and also by glucoand lipotoxicity on insulin target tissues leading to more severe insulin resistance. This decompensation phenomenon could explain why people with type 2 diabetes often present with severe hyperglycemia and why patients with pre-type 1 diabetes when stressed can develop severe hyperglycemia and then go into remission once the stress recedes. Perhaps the LPx and HPx rats represent different stages of diabetes development. We do not know if the LPx rats, followed for a longer period of time, would display the same changes in islet gene expression as HPx rats, which might lead to decompensation with increasing hyperglycemia. It is equally possible that the less severe phenotype would be stable and long lasting. Future experiments will be necessary to address this question.
Loss of ␤-Cell Differentiation in Px Rats-This study and others (13)(14)(15)26) provide evidence that ␤-cell maladaptation to the diabetic milieu is associated with a global disruption in ␤-cell gene expression rather than with specific gene defects as occur with maturity onset diabetes of the young 1-6 and mitochondrial diabetes. Thus, under the influence of diabetes, otherwise normal ␤-cells show progressively decreased expression of a panel of genes important for glucose-stimulated insu-

␤-Cell Abnormalities Worsen with Duration of Hyperglycemia
lin secretion and the regulation of ␤-cell gene expression. Genes that are highly expressed in ␤-cells such as insulin, GLUT2, glucokinase, mitochondrial glycerol phosphate dehydrogenase, pyruvate carboxylase, potassium, and calcium channels and several islet-associated transcription factors show a consistent 40 -60% decrease in mRNA levels in Px rats. On the other hand, concomitant up-regulation of several normally suppressed genes (LDH-A, hexokinase I, glucose-6-phosphatase, stress genes, and the transcription factor, c-Myc) appears to be an integral part of the change in ␤-cell phenotype with diabetes. Similarly, in models of diabetes derived by genetic modification such as the Zucker diabetic fatty (ZDF) rat, progression to diabetes is associated with a similar global alteration in ␤-cell gene expression (26).
Changes in Gene Expression and the Loss of GIIS after Px-The loss of GIIS has been identified in not only type 2 and early type 1 diabetes but also in islet transplants (27)(28)(29)(30)(31). Islets from Px rats display a similar defect in GIIS (16,29). The downregulation of islet-associated transcription factors (PDX-1, BE-TA2/NeuroD, HNF1␣, and Nkx6.1) after Px could contribute to the altered expression of genes essential for GIIS. None of the transcription factors were completely shut off in Px rats, but recent studies implicate the importance of modest reductions in the regulation of ␤-cell gene expression (32,33). Moreover, overexpression of PDX-1 can induce insulin gene expression in nonislet tissue, which can improve glucose homeostasis in diabetic mice (34). In Px rats, the down-regulation of these and other potentially important islet transcription factors (Pax6, PAN1, IB1, HNF3␤, HNF4␣1, HNF4␣2/5) paralleled the decreased expression of metabolic enzymes GLUT2, glucokinase, mitochondrial glycerol phosphate dehydrogenase, and pyruvate carboxylase. The expression of these genes favors the delivery of glucose metabolites to mitochondria and the generation of metabolic signals, such as ATP, that lead to an appropriate insulin secretory response to an extracellular glucose stimulus (2,5). Their decreased expression coupled with upregulation of normally suppressed metabolic genes may be sufficient to interfere with this unique and possibly fragile glucose recognition mechanism. In theory, the increased expression of genes such as glucose-6-phosphatase and LDH-A could up-regulate metabolic pathways diversionary to normal ␤-cell metabolism and thus impair the efficiency of carbon flux through glycolysis and the shuttling of molecules to mitochondria for oxidation and ATP formation (35,36). Normally, transported glucose is efficiently metabolized by glycolysis and mi-

␤-Cell Abnormalities Worsen with Duration of Hyperglycemia
tochondrial oxidation with little if any carbon converted to lactate (37,38). ␤-Cell Hypertrophy after Px-Hypertrophy can be a compensatory response to increased demand in terminally senescent cells. Hypertrophy of ␤-cells has been found in Px rats (13,17), prediabetic ZDF rats with impaired glucose tolerance (4), and in pregnancy (39). The up-regulation of c-Myc may be important in the compensatory growth of ␤-cells, since this factor can lead to hypertrophy in the absence of cell division (40). The development of ␤-cell hypertrophy is likely to be dependent upon the activation of cell cycle inhibitors such as p21 and/or survival factors that prevent c-Myc-induced proliferation and apoptosis (14,41,42). ␤-Cell hypertrophy was evident 14 weeks after Px and thus appears unrelated to the initial burst in regeneration of the endocrine pancreas that occurs in the first 7-10 days after surgery (18). Rather, the findings suggest that ␤-cell hypertrophy represents a stable response to the diabetic environment that, although inadequate, may prevent more serious metabolic decompensation.
Role of Hyperglycemia in the Loss of ␤-Cell Differentiation in Px Rats-Chronic hyperglycemia has been recognized as a leading cause of ␤-cell dysfunction (5,29,31). In humans with diabetes, any treatment that normalizes the plasma glucose profile leads to improvements in insulin secretion (43,44). Our findings suggest that the severity of hyperglycemia plays a critical role in the progressive loss of ␤-cell differentiation with diabetes. Circulating lipids were not altered in Px rats at 4 weeks or in LPx rats at 14 weeks, suggesting that they do not influence the changes in ␤-cell gene expression in the Px model. Other studies have shown that fatty acids can lead to alterations in ␤-cell gene expression (45)(46)(47)(48), and the changes in ZDF rats have been associated with elevated circulating fatty acids (47,49). However, a recent study in ZDF rats found hyperglycemia, and not hyperlidemia, to be associated with the decreased insulin gene expression in this model of diabetes (50). Therefore, ␤-cell dysfunction in Px and ZDF rats may be similarly mediated by chronic hyperglycemia. These findings do not rule out an important role for intracellular lipid pathways in the ␤-cell dysfunction of diabetes. Key deleterious changes could be caused by the effects of high glucose concentrations working in concert with free fatty acid substrate provided by normal levels of circulating free fatty acid.
Deterioration of the ␤-Cell Phenotype with Time-The duration of exposure to hyperglycemia appears a critical factor in the deterioration of the ␤-cell phenotype as indicated by the following: 1) short term hyperglycemia maintained by glucose clamp induced little change in ␤-cell gene expression (Table  IV), whereas long term hyperglycemia after Px induced global alterations (Table II), 2) changes in ␤-cell gene expression were fully reversible with phlorizin treatment of Px rats at 4 weeks (13-15) but not at 14 weeks (Fig. 5), and 3) low levels of hyperglycemia in Px rats induced significant changes in gene expression after 14 weeks but not at 4 weeks (Table II). These data establish a relationship between the duration of hyperglycemia and the deterioration of the ␤-cell phenotype with diabetes. It is noteworthy that at 4 and 14 weeks, rats with such similar minimal hyperglycemia have such striking differences in ␤-cell phenotype. The ␤-cell hypertrophy of this later time point is documented, although overall ␤-cell mass was not measured. These results raise important questions about the apparent ability of ␤-cells with markedly altered phenotype to maintain enough insulin output to keep glucose levels in the nearly normal range for long periods of time. There may be important lessons about how ␤-cells can adapt to prevent deterioration to frank diabetes. It is possible that the ␤-cell changes seen in this rat model resemble those of humans with the state of impaired glucose tolerance, a condition of mild hyperglycemia that can persist for many years before deterioration to the state of frank type 2 diabetes.
In conclusion, the findings of this study are consistent with our hypothesis that a critical loss of ␤-cell differentiation contributes to the ␤-cell dysfunction found in diabetes (5). However, despite a marked loss in ␤-cell phenotype, insulin output can be well enough maintained to keep glucose levels in a close to normal range.