Pancreatic beta-cell-specific targeted disruption of glucokinase gene. Diabetes mellitus due to defective insulin secretion to glucose.

Mice carrying a null mutation in the glucokinase (GK) gene in pancreatic beta-cells, but not in the liver, were generated by disrupting the beta-cell-specific exon. Heterozygous mutant mice showed early-onset mild diabetes due to impaired insulin-secretory response to glucose. Homozygotes showed severe diabetes shortly after birth and died within a week. GK-deficient islets isolated from homozygotes showed defective insulin secretion in response to glucose, while they responded to other secretagogues: almost normally to arginine and to some extent to sulfonylureas. These data provide the first direct proof that GK serves as a glucose sensor molecule for insulin secretion and plays a pivotal role in glucose homeostasis. GK-deficient mice serve as an animal model of the insulin-secretory defect in human non-insulin-dependent diabetes mellitus.

Mice carrying a null mutation in the glucokinase (GK) gene in pancreatic ␤-cells, but not in the liver, were generated by disrupting the ␤-cell-specific exon. Heterozygous mutant mice showed early-onset mild diabetes due to impaired insulin-secretory response to glucose. Homozygotes showed severe diabetes shortly after birth and died within a week. GK-deficient islets isolated from homozygotes showed defective insulin secretion in response to glucose, while they responded to other secretagogues: almost normally to arginine and to some extent to sulfonylureas. These data provide the first direct proof that GK serves as a glucose sensor molecule for insulin secretion and plays a pivotal role in glucose homeostasis. GK-deficient mice serve as an animal model of the insulin-secretory defect in human noninsulin-dependent diabetes mellitus.
Glucokinase (GK), 1 mainly expressed in pancreatic ␤-cells and the liver, is thought to constitute a rate-limiting step in glucose metabolism in these tissues (1)(2)(3)(4). Since insulin secretion parallels glucose metabolism and the high K m of GK (5-8 mM) ensures that it can change its enzymatic activity within the physiological range of glucose concentrations, GK has been proposed to act as a glucose sensor in the pancreatic ␤-cell (1,5). Recently, mutations of the GK gene have been identified in patients with maturity-onset diabetes of the young, a subtype of early-onset non-insulin-dependent diabetes mellitus (NIDDM) (6 -8). However, since all the mutations in humans so far occur in the region of the gene that is common to pancreatic ␤-cells and hepatocytes (9), and are heterozygous, it may not have been possible to fully reveal physiological roles of pancreatic ␤-cell GK either in vivo or in vitro. To this end, mice carrying a null mutation in the GK gene in pancreatic ␤-cells, but not in the liver, were generated by homologous recombination. Heterozygous mutant mice showed early-onset mild diabetes resembling the phenotype for human maturity-onset diabetes of the young. Homozygotes showed severe diabetes shortly after birth and died within a week. GK-deficient islets showed defective insulin secretion in response to glucose, while they responded to other secretagogues: almost normally to arginine and to some extent to sulfonylureas. These data provide the first direct proof that GK serves as a glucose sensor molecule for insulin secretion and plays a pivotal role in glucose homeostasis.

EXPERIMENTAL PROCEDURES
Cloning of the Mouse GK Gene, Construction of a Targeting Vector, and Homologous Recombinant Experiments-A DNA fragment including the pancreatic ␤-cell-specific exon 1␤ of the GK gene was cloned from a BALB/c mouse genomic library (Clontech). A BamHI site was introduced 30 base pairs 3Ј to the translation initiation codon of GK by the Kunkel method (10). A neomycin resistance gene (neo r ) with a pgk-1 promoter but without a poly(A) ϩ addition signal was substituted for the XbaI-BamHI fragment in the exon 1␤. A diphtheria toxin A fragment gene (DTA) with a MC1 promoter was ligated on the 3Ј terminus across the vector backbone, for negative selection (11,12). Homologous recombinant experiments in embryonic stem cells (ES cells) (A3-1) (13,14) were carried out as described previously (14). These cells were injected into blastocysts from C57BL/6J mice or co-cultured with morulae from C57BL/6J mice (15,16) and transferred into pseudopregnant ICR females to generate offspring.
Determination of Glucose Phosphorylating Activity-Islets were isolated from the pancreas of 3-7-day-old mice by collagenase digestion method (17). Glucose phosphorylating activities by hexokinase (HK) and GK were determined fluorometrically (18) except that GK activity was measured at 50 mM glucose.
Glucose Tolerance Test-Mice (10 weeks old) were fasted for more than 16 h before the study. They were then loaded with 1.5 mg g Ϫ1 (body weight) glucose by intraperitoneal infusion. Blood samples were taken at different time points from the orbital sinus. Insulin levels were determined using an insulin radioimmunoassay kit (Shionogi) with rat insulin as standard.
Insulin Content Assay-Isolated islets were suspended in 100 l of acid ethanol, and cellular insulin was extracted, diluted (100 times), and assayed by radioimmunoassay.
Determination of Intracellular Calcium Concentration-Isolated islets were incubated overnight with RPMI 1640 medium and measure-* This work was supported by Grant 192125 from the Juvenile Diabetes Foundation International (to T. K.) and by a grant for diabetes research from the Ohtsuka Pharmaceutical Co., Ltd. (to T. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Batch Incubation Study-Batches of 10 islets were preincubated for 60 min at 37°C in 5% CO 2 in 1 ml of Hank's-buffered saline containing 0.2% bovine serum albumin, 10 mM HEPES, pH 7.5 (Hanks'-BSA buffer) plus 0.1 mM glucose. The medium was then replaced with 1 ml of Hanks'-BSA buffer supplemented with secretagogues. After a 60-min incubation at 37°C, the medium was removed for radioimmunoassay of insulin (Shionogi).

RESULTS AND DISCUSSION
Targeted Disruption of Glucokinase Gene in Pancreatic ␤-Cells-Alternative splicing of a single GK gene gives rise to two isoforms of GK, one expressed in the pancreatic ␤-cells and the other in liver, which have different first exons (exon 1␤ and exon 1L, respectively) ( Fig. 1A) (9). These two isoforms are transcribed by two different promoters, and the downstream promoter, which lies between exon 1␤ and 1L, drives transcription of the liver GK isoform. We were therefore able to disrupt exon 1␤ expression and thereby selectively eliminate expression of the pancreatic ␤-cell isoform of GK without affecting expression of the liver isoform. Eight independent ES cell clones were identified as carrying the targeted mutant GK allele (Fig. 1B). Male chimeras originated from two homologous recombinant clones transmitted the mutant GK allele to their offspring. Heterozygous mutant mice were apparently normal and gave birth to mice homozygous for the mutant GK allele. The ratio of wild-type, heterozygous, and homozygous mice was 80:170:85 in 335 offsprings at 3-4 days of age, which was consistent with Mendelian inheritance (Fig. 1B). Reverse-transcriptase PCR analysis revealed that GK expression in pancreatic ␤-islets was completely absent in homozygotes (data not shown). In islets from homozygous neonates, GK activity was completely absent, whereas HK activity was similar to that from wild-type neonates (Fig. 1C). In islets from adult heterozygotes, GK activity (V max ) was 48% of that from the wildtype mice. In contrast to the ␤-cell, both GK and HK activities in the liver were unaltered in each genotype (data not shown).
Mild Diabetes in Heterozygous GK Knock-out Mice-At birth, the blood glucose levels of heterozygous and wild-type mice were 2.5 Ϯ 0.3 (n ϭ 5), and 2.4 Ϯ 0.1 (n ϭ 4) mM, respectively. However, about 50% of the heterozygous mice showed mild glycosuria within a day, suggesting the development of earlyonset diabetes mellitus. Heterozygous mice (10 weeks old) showed significantly higher blood glucose levels both before and after a glucose load and smaller increments in serum insulin levels than wild-type mice (Fig. 2). Insulin tolerance test revealed that the heterozygous mutant mice were as sensitive to insulin as the wild-type (data not shown). These results demonstrate that a heterozygous mutation of the GK gene in the pancreatic ␤-cells is sufficient to cause impaired insulin secretion to glucose and diabetes mellitus. In this respect, Efrat et al. (21) have generated mice in which pancreatic GK expression was attenuated by a ribozyme-mediated method. Although GK activity of these mice was only 30% that of normal and they showed an impaired insulin response to glucose in perfused pancreas experiments, the fasting and postprandial glucose levels remained normal. It is possible that the different strategies used to attenuate GK expression or the different strains of mice used (Efrat et al. used C3H, whereas this report used ICR) may explain the apparent discrepancy between their and our results.
Characterization of Homozygous GK Knock-out Mice-Ho- Diabetes Mellitus in Mice Lacking ␤-Cell Glucokinase 30254 mozygous mutant pups were normal in size, appearance, and body weight at birth (Fig. 3A). Their blood glucose level was 2.7 Ϯ 0.1 mM (n ϭ 4), which was indistinguishable from that of wild-type mice (2.4 Ϯ 0.1 mM, n ϭ 4). Although they were able to suck as evidenced by the presence of milk in their stomachs, they showed no increase in body weight with age (Fig. 3A). They also showed marked glycosuria within a day, and almost all animals died within 7 days of birth apparently due to dehydration. At 3-4 days of age, their blood glucose levels were markedly higher than those of wild-type or heterozygous animals, while serum insulin levels were low relative to the elevated blood glucose concentrations, suggesting the presence of relative insulin deficiency (Fig. 3B). Only 20% of homozygotes showed ketosis in spite of marked hyperglycemia. These suggest that basal insulin secretion is preserved in homozygotes presumably due to activity of the ␤-cell HK. Thus, diabetes mellitus in the GK-deficient mice, although severe, is more similar to NIDDM rather than insulin-dependent diabetes mellitus. Post-mortem investigation revealed no gross abnormalities in any of the organs examined except for occasional fatty changes of the liver in homozygous mutants (data not shown). 9 out of 10 homozygous mutant pups subcutaneously injected with 10 -20 milliunits of human insulin (Novolin R, Novo) twice a day survived beyond 7 days of age, and 4 out of 6 homozygous mutant pups orally administered with 20 mol of glibenclamide (kindly provided by Yamanouchi Pharmaceutical Co.) per day survived beyond 10 days of age, while all untreated pups (n ϭ 50) died before 7 days of age. Both of these treatments significantly lowered blood glucose levels (by 20 -40%) and caused a gain in body weight (to approximately 80% of the wild-type littermate). Since GK is also expressed in rare neu-roendocrine cells in the brain and gut (22), it is possible that the absence of glucose sensing in the brain or gut by GK may have modulated the phenotype of homozygous mutant pups. Nevertheless, it seems likely that hyperglycemia due to a lack of ␤-cell GK is the major cause of severe metabolic failure and early death in several days, since insulin or sulfonylurea improved these features. Immunostaining for glucagon, insulin, and somatostatin revealed that differentiation into pancreatic ␣, ␤, and ␦ cells also appeared to be unaffected (Fig. 4). Although there may be subtle changes in the architecture of ␣, ␤, and ␦ cells in the islets, its gross appearance was normal. The insulin contents of the wild-type, heterozygous, and homozygous GK-deficient islets were 1.7 Ϯ 0.3 (n ϭ 5), 2.2 Ϯ 0.7 (n ϭ 6), and 2.5 Ϯ 0.2 (n ϭ 10) ng of insulin/islet, respectively. These data indicate that ␤-cell GK is not essential for normal development, differentiation of endocrine pancreas, or insulin biosynthesis.
Characterization of GK-deficient Islets-Impact of lack of GK in the pancreatic ␤-cell was investigated in isolated islets from 7-10 days old mice. A rise in the intracellular Ca 2ϩ concentration in the ␤-cell ([Ca 2ϩ ] i ) is a key event in glucose-stimulated insulin secretion (19,20). Although all the GK-deficient islets looked alike under microscopy, they could be subdivided into two groups in terms of basal [Ca 2ϩ ] i levels; those with normal calcium levels (110 Ϯ 34 nM, n ϭ 7; about 20% of the islets), and those with higher basal calcium levels (higher than 200 nM; about 80%) which might reflect some damage of the islets. In the former group, the increase in [Ca 2ϩ ] i elicited by glucose was completely abolished, while the increase evoked by glibenclamide or arginine was essentially normal (Fig. 5A). However, in islets with higher basal calcium levels, the increase in [Ca 2ϩ ] i elicited not only by glucose but also by glibenclamide was completely abolished and that by arginine modestly impaired (data not shown). The observed heterogeneity of GKdeficient islets may be due to a direct effect of the impaired glucose metabolism in GK-deficient islets or may be a consequence of the hyperglycemia and other metabolic defects. Even in wild-type or heterozygous GK-deficient islets, there were islets with higher basal calcium levels (more than 200 nM), but Diabetes Mellitus in Mice Lacking ␤-Cell Glucokinase 30255 its proportion was less than 10%. We next examined insulin secretion using the batch incubation method (Fig. 5B). Insulin secretion from heterozygous GK-deficient islets in response to 0.1 mM or 3 mM glucose was normal, but that in response to 10 mM glucose was significantly impaired compared with wild-type islets. Impairment in insulin secretion in response to 20 mM glucose was less evident. On the other hand, insulin secretion in response to glibenclamide or arginine was unaffected. In homozygous GK-deficient islets, although there was some basal insulin secretion at 0.1 mM glucose, presumably due to the activity of HK, increase in insulin secretion in response to 20 mM glucose was completely abolished. In contrast, insulin secretion in response to arginine was essentially preserved. Regarding insulin secretion in response to glibenclamide, it was decreased by about 50 -80% depending on the method of estimation (Fig. 5B). Nevertheless, since the islets used here were supposed to be a mixture of those with normal and those with higher basal calcium levels (expected to be 20% and 80% of the population, respectively), which were responsive and unresponsive to glibenclamide in calcium study, we interpreted these results as suggesting that islets with normal basal calcium levels may have responded to sulfonylurea in insulin secretion.
The insulin secretory response to a physiological increment in glucose concentration was impaired in heterozygous GKdeficient islets and completely defective in homozygous GK-deficient islets despite the presence of HK, supporting the concept that GK serves as a glucose sensor molecule for insulin secretion. This is consistent with the smaller increments in serum insulin levels after a glucose load in heterozygous mice (Fig. 2), lack of increments in insulin levels in homozygous mice in spite of hyperglycemia (Fig. 3B), and the secretory abnormalities in human subjects with GK mutations (23). GK-deficient islets responded to non-glucose secretagogues in insulin secretion (almost normally to arginine and to some extent to sulfonylureas), indicating that GK is not absolutely required for insulin secretion in response to these secretagogues. It should also be noted that insulin secretion in response to glibenclamide was impaired, suggesting that GK may play an important role in insulin secretion in response to some of nonglucose secretagogues such as glibenclamide. This possibility would be examined in future. The heterozygous or insulintreated homozygous mutant mice described here provide the first animal model of diabetes with a defined genetic defect in insulin secretion, and should give important insights into the pathogenesis and development of human NIDDM.