In vivo mutagenesis of the insulin receptor.

Mice bearing targeted gene mutations that affect insulin receptor (Insr) function have contributed important new information on the pathogenesis of type 2 diabetes. Whereas complete Insr ablation is lethal, conditional mutagenesis in selected tissues has more limited consequences on metabolism. Studies of mice with tissue-specific ablation of Insr have indicated that both canonical (e.g. muscle and adipose tissue) and noncanonical (e.g. liver, pancreatic beta-cells, and brain) insulin target tissues can contribute to insulin resistance, albeit in a pathogenically distinct fashion. Furthermore, experimental crosses of Insr mutants with mice carrying mutations that affect insulin action at more distal steps of the insulin signaling cascade have begun to unravel the genetics of type 2 diabetes. These studies are consistent with an oligogenic inheritance, in which synergistic interactions among few alleles may account for the genetic susceptibility to diabetes. In addition to mutant alleles conferring an increased risk of diabetes, these studies have uncovered mutations that protect against insulin resistance, thus providing proof-of-principle for the notion that certain alleles may confer resistance to diabetes.

Diabetes is a growing threat to public health worldwide (1). Type 1 diabetes is caused by autoimmune destruction of pancreatic ␤-cells (2), whereas type 2 diabetes results from insulin resistance and impaired ␤-cell function (3). Insulin resistance is found in the main insulin target tissues (muscle, adipose cells, liver) of patients with overt diabetes. However, this is a consequence of chronic hyperinsulinemia and glucotoxicity (4). The question of whether insulin resistance represents a generalized impairment of insulin action or is initially restricted to specific organs has remained unclear, as has its relationship to impaired ␤-cell function. Although insulin receptor (Insr) 1 defects are uncommon as a cause of diabetes (5), this gene remains an attractive target for in vivo studies of insulin resistance, as it has been shown to be the master switch of the metabolic (6, 7) and growth-promoting actions (8,9) of insulin.

Insulin Receptor Gene Knock-out
Mice homozygous for null Insr alleles are born at term with slight growth retardation (9) but rapidly develop metabolic abnormalities, followed by diabetic ketoacidosis and death (6,7). The marked difference between this phenotype and that of humans lacking INSR (5) has been reviewed elsewhere (10).

Conditional Insr Ablation
The lethal phenotype of Insr knock-out mice precludes a detailed analysis of Insr function in different tissues in adult mice. This problem has been circumvented by generating conditional knockouts using the Cre/loxP binary system (11) or combined haploinsufficient and dominant-negative mutations (Table I).

Insulin Action in Skeletal Muscle and
Insulin Resistance The cornerstone of current theories on the pathogenesis of type 2 diabetes is that skeletal muscle, the main site of insulin-dependent glucose disposal, is intrinsically unable to respond to insulin, either as a result of genetic predisposition or as a consequence of environmental factors (4). Indeed, an impairment of insulin-dependent glucose uptake and phosphorylation is an early step in the development of type 2 diabetes (12). Thus, several studies have tried to replicate this abnormality using either dominant-negative mutations (13,14) or conditional inactivation of Insr (15) to abrogate insulin signaling. When Insr was inactivated by Cre-mediated recombination, MIRKO mice developed a metabolic syndrome with increased fat stores, hypertriglyceridemia (15). Nevertheless, they failed to develop hyperinsulinemia and diabetes, in part because they are able to shunt glucose utilization from muscle to adipose tissue (16) (Fig. 1). Similarly, in crosses of mice heterozygous for a systemic Insr knock-out with transgenics bearing a dominant-negative Insr transgene in muscle, there was a greater than 90% decrease in Insr kinase activity and a blunting of insulin-dependent glucose uptake but no diabetes (14). Results in transgenics expressing the dominant-negative transgene were inconclusive (13).
These observations are unexpected because the prediction was that impaired insulin signaling in muscle would lead to generalized insulin resistance. The findings stand in sharp contrast to studies showing that ablation of the insulin-dependent glucose transporter Glut4 in skeletal muscle can cause diabetes (17). There are several explanations for this apparent conundrum. In mice lacking muscle Insr, two pathways can compensate for the ablation of insulin signaling: the Igf1r pathway (18) and the contraction-activated pathway (19,20).
The importance of Igf1r in muscle metabolism is highlighted by a mouse model of combined ablation of insulin and Igf1 receptor function in skeletal muscle, using a dominant-negative Igf1r transgene. The mutant Igf1r impairs Insr function through trans-dominant inhibition of the kinase activity of heterodimers composed of an Insr monomer and a mutant Igf1r monomer. Unlike mice with isolated Insr ablation, these mice do develop diabetes with all the characteristic changes of the insulin-resistant state (21).
Collectively, these data indicate that there are branching pathways leading to glucose uptake and Glut4 translocation in skeletal muscle and that these compensatory mechanisms enable mice lacking Insr to overcome the impairment of insulin signaling. In contrast, when Glut4 is mutated, the impairment of glucose transport in muscle can result in severe metabolic derangement. It should be emphasized, however, that not all Glut4 muscle-specific knock-out mice develop diabetes. This may be due to genetic variability among different mice on an outbred background. Whereas these models confirm the paramount role of skeletal muscle as a site of insulin action, they also highlight the role of genetic modifiers in determining muscle insulin sensitivity (see below).

Adipose Tissue
Conditional ablation of Insr in adipocytes has been used to address how insulin signaling affects the development of the common metabolic complications arising from obesity. When Insr was ablated in white and brown adipocytes using a Cre transgene driven by the adipose-specific aP2 promoter (FIRKO) (22), mice showed an ϳ50% decrease of gonadal fat mass and whole body triglyceride content. Moreover, FIRKO mice are resistant to gaining weight during aging or following administration of gold-thioglucose, a hypothalamic toxin that leads to hyperphagia and obesity in normal mice. Similarly, FIRKO mice are protected against hyperphagia-induced glucose intolerance. These findings indicate that insulin signaling in adipose cells is not critical for the maintenance of euglycemia in mice but is required for triglyceride storage in adipocytes. The observation that insulin-dependent glucose uptake is nearly absent in FIRKO mice, whereas triglycerides are only 50% lower than normal, raises the question as to the source of 3-glycerol phosphate to carry out triglyceride synthesis. One potential pathway is glycolysis from insulin-independent glucose uptake through Glut1. Another pathway, reviewed elsewhere in this series, is glyceroneogenesis, i.e. the generation of 3-glycerol phosphate from pyruvate, lactate, and amino acids mediated by PEPCK (23). It is conceivable that the absence of Insr signaling increases PEPCK activity, leading to increased intracellular re-esterification of triglycerides as a mechanism to preserve scarce 3-glycerol phosphate derived from glycolysis.
As with muscle, knock-out of the insulin-dependent glucose transporter Glut4 in fat has more profound effects on glucose homeostasis than Insr knock-out, indicating that mutations at different steps in the insulin action cascade can differ in their impact on whole body response to insulin (24). The metabolic changes in FIRKO are accompanied by a redistribution of adipocyte size, in which fat pads have a decreased content of intermediatesized adipose cells, with an increase of large and small cells. These data will likely contribute to rekindling the debate regarding the complex relationship between adipocyte size and insulin sensitivity. In addition to its role in mature adipocytes, Insr appears to play a pivotal developmental role in adipogenesis. Targeted Insr inactivation in 3T3-L1 cells impairs their ability to fully differentiate into adipocytes (25).
Another interesting phenotype associated with impaired insulin receptor signaling in adipocytes is the increase in longevity. FIRKO mice have an ϳ20% increase in mean, median, and maximum lifespans. These data support the notion that a decreased fat mass can affect lifespan independently of caloric restriction (27) and should be viewed in the context of the life-prolonging effects of mutations affecting insulin/Igf signaling in Caenorhabditis elegans (28 -32).
Insr has been inactivated in brown adipose tissue using transgenic mice in which the uncoupling protein-1 gene promoter was used to drive selective ablation of the "floxed" Insr locus (BAT-IRKO). These mice display an age-dependent loss of brown adipose tissue, accompanied by a deterioration of ␤-cell function and a decrease of ␤-cell mass, giving rise to hyperglycemia (33). More studies are required to examine the interaction between brown adipose tissue and pancreatic ␤-cells.

The Role of Non-canonical Insulin Target Tissues in
Insulin Resistance Conditional mutagenesis of Insr has been especially valuable in addressing the controversial question of whether direct or indirect effects of insulin are predominant in tissues other than muscle and fat. Simply put, the question is whether insulin regulation of complex responses, such as glucose production in the liver, appetite regulation in the brain, and gonadotropin response in the ovary, is a direct result of the activation of Insr pathways or is a secondary effect of insulin-induced substrate redistribution. Because the current canon is that insulin resistance affects primarily tissues that have the ability to respond to insulin by increasing glucose uptake, such as muscle and adipose cells, we refer to the other target tissues of insulin as "non-canonical." Nowhere has this controversy been more keenly felt than in studies of insulin regulation of hepatic glucose production. To simplify a large body of rather complex evidence, the question is whether insulin suppression of hepatic glucose production, the failure of which causes fasting hyperglycemia, is mainly a result of Insr signaling in hepatocytes or of a reduced flux of gluconeogenic precursors and free fatty acids from muscle and adipose tissue, as well as inhibition of glucagon secretion (34). Before we review the work addressing the role of hepatic Insr signaling in this process, we should emphasize that a wholesale application of these lessons to human metabolism would be misleading, as patterns of tissue glycogen storage in rodents are different from those found in other species. For example, in humans the amount of glycogen per g of tissue protein in liver and muscle is comparable, whereas in rodents hepatic glycogen is 10-fold or more abundant than muscle glycogen. This important qualifier is all too often overlooked.
Initial evidence that the effect of insulin on hepatic glucose production is largely a direct consequence of insulin binding to its receptor is derived from mice in which Insr is ablated in muscle and adipose tissue, with normal insulin signaling in the liver (14). These mice develop impaired glucose tolerance, do not progress to diabetes, and maintain normal hepatic insulin sensitivity. The data provide indirect evidence that hepatic insulin resistance is required for the onset of overt diabetes. However, they also suggest that insulin resistance in the liver is not merely a by-product of insulin resistance elsewhere but rather an intrinsic abnormality of insulin signaling in hepatocytes.  1. Synopsis of conditional mutagenesis of Insr. As complete inactivation of Insr is lethal immediately after birth, methods have been developed to ablate Insr function in selected tissues using Cre/loxP-mediated recombination, as well as dominant-negative transgenes. The results of these experiments are summarized in this figure. For each tissue, we report the salient features of the relevant Insr knock-out. References to each knockout are provided in Table I and throughout the text.

Minireview: Tissue-specific Insulin Resistance 28360
This prediction found experimental support in mice with conditional, liver-specific Insr knock-out (LIRKO). LIRKO mice exhibit marked insulin resistance, glucose intolerance, and a failure of insulin to suppress hepatic glucose production and to regulate hepatic gene expression (35). In addition, the LIRKO mouse exhibits marked hyperinsulinemia, with a 50% reduction in circulating triglycerides and a trend toward lower free fatty acid levels. These data indicate a critical role for hepatic Insr in regulating glucose homeostasis, insulin clearance and hepatocyte lipid synthesis.
Although the interpretation of the LIRKO phenotype is complicated by the onset of liver failure with age, these mice represent an important reagent for addressing the role of hepatic insulin action in the pathogenesis of type 2 diabetes. Measurements of hepatic glucose fluxes using hyperinsulinemic-euglycemic clamps reveal that insulin fails to suppress gluconeogenesis in LIRKO mice, consistent with the view that Insr signaling is required for both indirect and direct effects of insulin on the liver in mice (36).

Insr Ablation in Neurons
Insr is widely expressed in several brain areas (37). Brain Insr has been implicated in the regulation of satiety, whereas glucose disposal occurs in an insulin-independent manner (38). This potential role of Insr has been studied by generating a neuron-specific Insr knock-out (NIRKO) using the nestin promoter. Ablation of Insr in nestin-positive neurons results in increased food intake and moderate diet-dependent obesity (39). The role of brain Insr on metabolism has also been studied using intra-cerebroventricular injections of antisense oligonucleotides and blocking antibodies to Insr (40). This manipulation impaired hypothalamic Insr function in rats, causing a rapid onset of hyperphagia and an increased fat mass, similar to the NIRKO mouse. In addition, using an insulinclamp, the authors showed that the ability of insulin to blunt hepatic glucose output was decreased by ϳ50%, extending the role of hypothalamic Insr to control of peripheral glucose disposal (41).
In addition to its metabolic functions, brain Insr appears to regulate gonadotropin production. In fact, NIRKO mice also develop hypogonadotropic hypogonadism, associated with impaired maturation of ovarian follicles in females and reduced spermatogenesis in males, leading to reduced fertility (39).

Insr Signaling and Pancreatic ␤-Cell Function
There is an ample body of literature on the role of receptor tyrosine kinase signaling in pancreatic ␤-cell proliferation and insulin secretion (42). For example, ablations of different Irs proteins have selective effects on insulin secretion (Irs1) (43) or ␤-cell proliferation (Irs2) (44).
The effects of targeted disruption of the three receptors of the Insr subfamily have been examined in ␤-cells from mice bearing either conditional or ubiquitous mutations of the relevant genes. The Insr-related receptor (Irr) is an orphan receptor belonging to this subfamily and is expressed at higher levels in ␤-cells than either insulin or Igf1 receptors (45). Metabolic analyses and insulin release studies from islets of Irr knock-outs have thus far failed to demonstrate a role for this receptor in ␤-cell function (46).
Insr has been inactivated in ␤-cells using an insulin 2 promoterdriven Cre transgene to obtain conditional recombination (␤IRKO). Lack of Insr in ␤-cells results in a selective impairment of glucosedependent insulin release and, in some mice, overt diabetes (47). This surprising observation raises the intriguing possibility that the two fundamental defects in type 2 diabetes, insulin resistance and ␤-cell failure, share a common pathogenesis.
Similarly, Igf1r ablation in ␤-cells results in impaired insulin secretion and altered glucose tolerance, without overt diabetes (48,49). Interestingly, the ultrastructure of ␤-cells lacking Insr or Igf1r is quite different. Whereas Insr-deficient ␤-cells do not display structural abnormalities (47), Igf1r-deficient ␤-cells are depleted of insulin secretory granules and enriched in Golgi stacks (49). These data could be construed to indicate that lack of Igf1r results in unregulated, constitutive insulin release, consistent with the known role of IGF1 in inhibiting insulin secretion in vivo (50).
The role of Insr and Igf1r in the pancreas is not limited to the endocrine function of the organ. Genetic epistasis experiments have recently revealed a novel function for the two receptors during pancreas development. Combined inactivation of Insr and Igf1r, but not of either receptor alone, resulted in a severe impairment of exocrine pancreatic development with preserved endocrine cell development (51). Because of the known ability of IGF2 to bind both Insr and Igf1r with equal affinity (52), the inhibition of pancreatic development in embryos lacking both receptors, but not either receptor alone, indicates that exocrine pancreatic growth is dependent on IGF2.
Notably absent from this array of phenotypes due to Insr and Igf1r ablation in ␤-cells are defects in cell proliferation, which have been shown to arise from ablation of Irs2. A simple explanation is that the two receptors can substitute for one another in promoting ␤-cell growth. This possibility will be addressed when double conditional mutants lacking both Insr and Igf1r are studied. Alternatively, it is possible that Irs2 mediates the actions of additional receptors in promoting ␤-cell proliferation. We have proposed that insulin/IGF signaling is important for proliferation and/or terminal differentiation of the elusive ␤-cell progenitor, presumably arising from cells embedded within pancreatic ducts (53).

Modeling the Genetics of Type 2 Diabetes in Insr Mutant Mice
We have also used the Insr heterozygous mutants to mimic genetic interactions leading to type 2 diabetes (54). A first step in this direction was the development of a polygenic model of insulinresistant diabetes by generating mice with combined heterozygous Insr and Irs1 mutations. Whereas Irs1 heterozygotes are normal, double heterozygous mice for both Insr-and Irs1-null alleles develop severe hyperinsulinemia and hyperplasia of pancreatic ␤-cells, and by 4 -6 months of age, nearly one-half of these mice become frankly hyperglycemic. This process closely resembles the pathogenesis of human diabetes (55,56). Thus, even a major predisposing allele, such as the null Insr mutation, has a modest effect by itself but plays a major role in the context of a predisposing background. This is confirmed by the observation that, when the double heterozygous knock-out mice are bred onto different genetic backgrounds, the prevalence of diabetes can vary from Ͻ2% to 85% (57). 2 The risk of the recurrence of diabetes in double heterozygous offspring (Insr/Irs1 ϩ/Ϫ ) of single heterozygous Insr parents increases 4-fold, similar to the excess risk of diabetes in first degree relatives of humans with diabetes (58,59). The findings in the Insr/Irs-1 ϩ/Ϫ mouse are consistent with an oligogenic mode of inheritance of type 2 diabetes, in which two subclinical defects of gene function can account for virtually the entire genetic susceptibility to the disease.
Whereas combined mutations of Insr and Irs1 have provided insight into the polygenic nature of type 2 diabetes, comparisons of double mutant mice with Insr/Irs1-or Insr/Irs2-null alleles provide an illustration of genetic heterogeneity, i.e. of the ability of different mutant loci to give rise to similar phenotypes. Insr/Irs1 and Insr/Irs2 double mutant mice develop diabetes with similar frequencies. However, Insr/Irs1 mice are primarily insulin-resistant in skeletal muscle, whereas the Insr/Irs2 are primarily insulinresistant in liver (55). These data indicate that different insulin receptor substrates play tissue-specific roles, with Irs1 being the primary mediator of insulin action in muscle and Irs2 in liver (60).
Not every mutation in the insulin signaling pathway is detrimental. For example, mutations of the regulatory p85 subunit of PI 3-kinase increase insulin sensitivity (61,62), and heterozygous p85 mutations protect from insulin resistance and diabetes caused by either Insr or Irs1 mutations by improving the efficiency of insulin signaling (63). Increased insulin sensitivity and resistance to dietinduced metabolic abnormalities is also observed in mice lacking the tyrosine phosphatase Ptp1b (64).
Likewise, haploinsufficiency for the forkhead transcription factor Foxo1 restores insulin sensitivity and prevents diabetes in Insr ϩ/Ϫ mice by decreasing expression of glucogenetic genes in liver, improving ␤-cell compensation, and decreasing adipocyte size (65). Foxo1 was identified as a potential negative regulator of insulin signaling based on studies in C. elegans (66). Indeed, transgenic mice expressing gain-of-function Foxo1 mutants are diabetic and insulin-resistant (65). Foxo1 appears to play a widespread role in the transcriptional response to insulin. Its targets include glu-2 C. R. Kahn, personal communication.

Conclusions
As we have pointed out in a recent publication (69), the contribution of Insr signaling to metabolic control appears to have been overestimated in canonical insulin target tissues, such as muscle and fat, and underestimated in non-canonical target tissues, such as liver, brain, and pancreatic ␤-cells. The findings in Insr mutant mice provide a better understanding of the protean manifestations of insulin resistance, expand the repertoire of potential targets for drug development, and suggest that treatments to improve insulin resistance should selectively modulate specific insulin responses in different tissues.