GLUT4 Gene Regulation and Manipulation*

A decade has passed since the cloning of the insulin-responsive glucose transporter, GLUT4. Numerous studies have demonstrated the complex hormonal and metabolic regulation of GLUT4 gene expression in adipose tissue and muscle. Careful dissection of the regulatory elements in the GLUT4 promoter has provided insight into the intricate control of this central gene of glucose homeostasis. Genetic manipulation of mice has provided further insight into the role of GLUT4 in carbohydrate and lipid metabolism at the whole body and tissue-specific levels. Analysis of GLUT41/2, GLUT4 null, and muscle-complemented GLUT4 knockout mice has furthered our understanding of peripheral insulin sensitivity. Additional studies on GLUT4 gene regulation and GLUT4 knockout models are likely to lead to novel therapies for type II diabetes and other diseases of insulin resistance.

the mRNA (5,6). Transcription rates using nuclear run-on transcription assays have demonstrated that the GLUT4 mRNA transcription rate is decreased in both adipose tissue and skeletal muscle in STZ-induced diabetic animals (5,6), whereas it is increased in skeletal muscle of fasted animals (6). Thus, changes in GLUT4 mRNA steady state levels reflect changes in the rate of mRNA synthesis.
The molecular basis for regulation of GLUT4 gene expression in states of relative insulin deficiency in vivo has been very difficult to solve. Insulin deficiency in vivo is complicated by the fact that compensatory counter-regulatory hormones are elevated. In addition, insulin deficiency is tightly coupled to plasma glucose levels and intracellular glucose utilization. For instance, STZ-diabetic animals are hyperglycemic and insulinopenic whereas fasted animals are hypoglycemic and insulinopenic, suggesting that insulin rather than circulating glucose levels are responsible for regulation of adipose tissue GLUT4 expression. This hypothesis was supported in studies using phlorizin to increase urinary output of glucose in diabetic rats (2). In contrast to insulin, phlorizin-induced normalization of glycemia in these insulin-deficient animals was unable to restore GLUT4 mRNA expression in adipose tissue. On the other hand, skeletal muscle GLUT4 mRNA is not down-regulated by insulinopenia associated with fasting, implying that insulin levels do not directly regulate GLUT4 gene expression (4).
Both chronically fasted and STZ-diabetic animals represent states of insulin deficiency where peripheral glucose metabolism is inhibited. With the production of transgenic mice overexpressing the GLUT4 gene, a model of insulinopenia in which peripheral glucose utilization is enhanced has been made available. Overexpression of human GLUT4 protein markedly enhanced glucose uptake and utilization in the fed state resulting in hypoglycemia and hypoinsulinemia (7). Despite the relative hypoinsulinemia, expression of the endogenous mouse GLUT4 mRNA was unaffected by the presence of the human GLUT4 protein (8). 2 This suggests the predominate metabolic control of GLUT4 gene expression is linked to intracellular glucose metabolism. The divergent effect of hypoinsulinemia of fasting compared with STZ-diabetes in skeletal muscle may be linked to differences in energy metabolism that occur in muscle in these different states. In vitro models for studying GLUT4 expression are limited by the small number of cultured cell models that express the GLUT4 gene. Differentiated murine 3T3-L1 and F442A adipocytes express relatively high levels of GLUT4 mRNA and protein similar to that found in primary isolated adipocytes or adipose tissue (9 -11). The appearance of GLUT4 mRNA in these cells first occurs about 4 days after the onset of differentiation (9). Unlike primary adipocytes or adipose tissue, 3T3-L1 adipocytes also express high levels of the GLUT1 glucose transporter isoform (10). These cell lines have been used as in vitro models to study several aspects of glucose transporter regulation, including regulation of gene expression. Interestingly, chronic exposure of 3T3-L1 in vitro cells or adipose tissue in vivo to insulin has differential effects on GLUT4 gene expression. Animals chronically treated with insulin show increased GLUT4 mRNA in adipose tissue (12,13) whereas chronic insulin treatment of 3T3-L1 adipocytes has resulted in either no change or in a marked reduction in GLUT4 mRNA levels (14,15). The different responses of GLUT4 mRNA to chronic insulin treatment in vivo and in vitro suggest GLUT4 mRNA does not respond directly to insulin action on adipose tissue. On the other hand, incubation of 3T3-L1 adipocytes in glucose-free medium down-regulates GLUT4 mRNA about 10-fold and is accompanied by the up-regulation of GLUT1 mRNA (16). Re-addition of glucose to the starved adipocytes restored GLUT4 mRNA levels. Supplementation of glucose-free medium with either fructose or pyruvate as an alternative energy source maintained the steady state level of GLUT4 mRNA. These data are * This minireview will be reprinted in the 1999 Minireview Compendium, which will be available in December, 1999. This is the third article of three in the "Insulin-stimulated Glucose Transport Minireview Series." This work was supported by National Institutes of Health Grants DK47425 and HL58119 (to M. J. C.) and DK47894 (to A. L. O.) and by grants from the American Diabetes Association (to M. J. C. and A. L. O.) and the American Heart Association (to M. J. C. consistent with a metabolic rather than hormonal regulation of GLUT4 gene expression.

Transcriptional Regulation of the GLUT4 Promoter
To understand how the factors described above exert their influence on GLUT4 gene expression, it is necessary to identify the molecular elements (cis-DNA sequences and trans-acting factors) responsible for the coordinate transcriptional control of gene expression. In the case of GLUT4, a complex pattern of gene expression is observed in various physiologic states that are difficult to mimic in vitro using traditional methods of studying the function of gene promotors in model cell culture systems. Without a clear understanding of the molecular basis for GLUT4 gene expression, it is difficult to develop a suitable in vitro model to understand how the GLUT4 gene is regulated in vivo. To circumvent these difficulties, promoter analysis using human GLUT4 reporter genes was performed in transgenic mice. This system allows analysis of transcriptional activity of the GLUT4 promoter in a natural physiologic context, which is essential for a gene that is subjected to a complex mixture of tissue-specific and nutritional/metabolic factors.
To date, 12 different transgenic constructs have been analyzed for appropriate tissue-specific and hormone-dependent GLUT4 gene regulation (Fig. 1). The first established transgenic line was engineered to express a human GLUT4 minigene consisting of the entire coding region of the gene and 5.3 kb of 5Ј-flanking DNA (8). Expression of this construct demonstrated that the human gene, in a mouse background, was subject to the same regulation as the mouse GLUT4 gene. Furthermore, the complex pattern of human GLUT4 transcription initiation site selection was also observed in these transgenic mice. A second line of transgenic mice carrying a construct in which expression of the chloramphenicol acetyltransferase (CAT) reporter gene was driven by 2.4 kb of 5Ј-flanking DNA demonstrated that this region of DNA was sufficient to confer not only tissue-specific expression but also regulated expression of the GLUT4 gene in chronic fasted and STZ-induced diabetic mice (8,17).
Because all the apparent regulatory cis-DNA elements were located within 2.4 kb of the transcription initiation site, studies have been concentrated on this area of the human GLUT4 promoter to define the functional elements that drive transcription. The initial approach to studying this promoter was to narrowly define the cis-acting DNA sequences required to support a full program of gene expression by generating CAT fusion genes driven by various fragments of the 5Ј-GLUT4 regulatory region. Using this approach the structure of the GLUT4 promoter and its regulatory regions has begun to be defined (Fig. 1).
A series of deletions in the 5Ј-end defined the first 900 base pairs upstream of the major transcription initiation site of the human GLUT4 gene as putative regulatory DNA (18). A comparison of sequences of this regulatory region in the human gene with the analogous regions of the rat and mouse GLUT4 gene revealed the existence of two areas with greater than 90% sequence identity referred to as domains I and II (19). A 5Ј deletion that removes both of these domains left a basal promoter which was ubiquitously expressed at a very low level in all tissues, including those that do not normally express the GLUT4 gene (20). A construct in which most of domain I was deleted leaving domain II intact was able to support a high level of CAT activity in skeletal muscle, but a low level of CAT expression was observed in heart and adipose tissue and tissues that do not normally express the GLUT4 gene (20). In this construct, domain II was not sufficient to support regulated expression of transgenic mRNA in STZ deficiency in skeletal muscle or any other tissue. Although domain II was not sufficient to support full transgenic expression in heart and adipose tissue, deletion of this region ablated transgene expression in these tissues. Thus domain II is necessary, but not sufficient, for full promoter function of the GLUT4 gene (18). Inspection of domain II revealed the existence of a perfectly conserved binding domain for the MEF2 family of transcription factors. A loss of function of this MEF2 binding domain had the same effect as deletion of the entire conserved region, demonstrating that this binding domain was the functional element within these sequences (18).
A DNA binding site functions by binding a specific protein complex. The MEF2 DNA binding site is known to bind isoforms of the MEF2 family (MEF2A, MEF2B, MEF2C, and MEF2D) of DNA binding proteins belonging to a larger family of MADS-box domain transcription factors (21). Although these transcription factors have been studied largely in the context of myogenesis, the expression of these proteins extends beyond muscle tissues. Using isoform-specific antibodies in electrophoretic mobility shift assays, MEF2A and MEF2D isoforms were shown to bind the GLUT4 MEF2 binding domain in skeletal muscle, heart, and adipose tissue (18). 2 These studies were the first to establish a role for MEF2 binding activity in gene expression in adipose tissue. Such a role is not unexpected as adipose tissue and skeletal muscle arise from embryonic mesoderm.
The identification and characterization of domain I in the human GLUT4 promoter are currently under way. Interestingly, analysis of the mouse GLUT4 promoter in 3T3-L1 adipocytes revealed an element responsible for insulin-mediated down-regulation of GLUT4 gene expression in those cells (22). This insulin-responsive element in mouse coincides with domain I in the human gene. Although insulin effects in vivo and in vitro are dissimilar, it remains possible that the insulin-responsive element defined in 3T3-L1 cells may be functional in mediating the effects of insulin deficiency in vivo. Establishing functional region(s) of this domain will pave the way for solving the complex regulation of the GLUT4 gene. Understanding the molecular basis for expression of the GLUT4 gene will be useful for targeting the expression of this gene in a manner appropriate for treatment of insulin-resistant glucose transport.

GLUT4؉/؊ Mice Exhibit Age-related Insulin Resistance and Diabetic Pathologies
Knockout mice with one null allele of GLUT4 (GLUT4ϩ/Ϫ) have been generated that exhibit reduced GLUT4 expression (23). Though all mice have the same mutation, namely one disrupted allele of GLUT4, a spectrum of phenotypes is observed with age. Specifically, when considering serum-fed glucose and insulin levels, GLUT4ϩ/Ϫ mice could be divided into three distinct groups: normal glycemia with normal insulin levels (N/N), normal glycemia with high insulin levels (N/H), or hyperglycemic with high insulin levels (H/H). The diabetes present in GLUT4ϩ/Ϫ mice is not coincident with alterations in body weight or fat pad weight. Curiously, adipocytes from diabetic H/H GLUT4ϩ/Ϫ mice are significantly enlarged. Secretion of cytokines such as tumor necrosis factor ␣ and leptin that influence insulin sensitivity and energy balance is altered in obese adipocytes (24), and this may contribute to the diabetic phenotype of the H/H mice. Additionally, neither pancreatic insufficiency, dyslipidemia, nor hepatic insulin resist- ance is associated with the GLUT4ϩ/Ϫ phenotype. Diabetic H/H GLUT4ϩ/Ϫ mice display a 40% increase in peak arterial blood pressure but a normal response to inotropic challenge, demonstrating that ventricular performance was not compromised. However, histopathological analysis revealed diabetic cardiomyopathies including cardiomyocyte hypertrophy, focal cell necrosis, interstitial fibrosis, and vascular sclerosis. Oddly, despite the lack of dyslipidemia in H/H GLUT4ϩ/Ϫ mice, micro-and macro-steatosis similar to that seen in diabetics was noted throughout the liver.
Beginning at 2 months of age all male GLUT4ϩ/Ϫ mice have reduced adipose and muscle GLUT4 expression (75 and 25-46%, respectively) (23). Decreased adipocyte GLUT4 content in GLUT4ϩ/Ϫ mice is the first measurable cellular defect to occur, the significance of which is yet to be realized. In vitro insulinstimulated glucose uptake into soleus and extensor digitorum longus (EDL) muscles of prediabetic N/H GLUT4ϩ/Ϫ mice is significantly diminished (23). Using euglycemic/hyperinsulinemic clamps it was demonstrated that the in vivo peripheral insulin resistance of prediabetic (N/H) GLUT4ϩ/Ϫ mice was as severe as in uncontrolled type II diabetes (25). Additionally, the rate of insulin-stimulated glycogen synthesis was significantly impaired directly because of reduced muscle GLUT4-mediated glucose uptake and not defective activation of glycogen synthase. The selective primary reduction in GLUT4 protein is the most direct proof to date of the rate-limiting role of glucose transport in whole body and skeletal muscle insulin-stimulated glucose uptake. This supports the idea that spontaneous variations in GLUT4 gene expression in human muscle may be a major determinant of peripheral insulin sensitivity. In vivo tracer studies also demonstrated that prediabetic N/H GLUT4ϩ/Ϫ mice exhibit a modest reduction in postabsorptive hepatic glucose production and normal intrahepatic distribution of glucose fluxes through glucose cycling, gluconeogenesis, and glycogenolysis (25). These data confirm the relatively minor role GLUT4 plays in basal glucose homeostasis.

Prevention of Insulin Resistance in GLUT4؉/؊ Expressing a Muscle-specific GLUT4 Transgene
Myosin light chain 1-GLUT4 (MLC-GLUT4) transgenic mice that specifically overexpress GLUT4 in fast twitch muscle (26) were mated into the genetic background of the GLUT4ϩ/Ϫ mutation to assess the therapeutic merit of muscle GLUT4 gene therapy in type II diabetes (23). 3 MLC-GLUT4 transgenic mice have increased insulin sensitivity without hypoglycemia (26). GLUT4 content and insulin-stimulated glucose uptake were normalized in fast twitch muscles of MLC-GLUT4ϩ/Ϫ mice. Fed plasma glucose and insulin levels were normal throughout the lives of MLC-GLUT4ϩ/Ϫ mice, and cardiac histopathologies were minimal. In vivo tracer studies demonstrated that whole body glucose utilization, glycolysis, and glycogen synthesis were normal in MLC-GLUT4ϩ/Ϫ mice, confirming the central role of muscle GLUT4 in peripheral insulin sensitivity and the genesis of diabetic complications. Although these studies strongly suggest muscle GLUT4 gene therapy could be effective in treating type II diabetes or other diseases of peripheral insulin resistance, they do not prove that point. To specifically test this hypothesis, design of an inducible, muscle-specific GLUT4 transgene (iGLUT4) is required (27). The iGLUT4 transgenically expressed in GLUT4ϩ/Ϫ mice could be activated at various points in the disease (e.g. N/H prediabetic; H/H diabetic) to determine whether muscle GLUT4 gene therapy can halt or reverse any or all diabetic complications.

GLUT4 Null Mice Maintain Normal Glycemia without GLUT4
In comparison with GLUT4ϩ/Ϫ mice it is both exciting and perplexing that mice which lack the insulin-sensitive glucose transporter (GLUT4 null) are not diabetic but do exhibit abnormalities in glucose and lipid metabolism (1, 28 -30). Surprisingly, blood glucose levels in GLUT4 null mice are normal under fasted and fed conditions. Although GLUT4 null mice have normal glucose tolerance, they do exhibit hyperinsulinemia in the fed state and impaired insulin tolerance, suggesting insulin resistance. Careful analysis of serum metabolites of GLUT4 null mice reveals a significant reduction in fed lactate and free fatty acid levels. Additionally, 9-fold reductions in fasting ketones are noted. GLUT4 null mice are 15-20% growth-retarded and have severely reduced adipose tissue depots and extreme cardiac hypertrophy. Northern and Western blot analyses verified that GLUT4 null mice could compensate for the lack of GLUT4 and maintain normal circulating glucose levels by a mechanism that did not involve overexpression of a known facilitative or Na ϩ -dependent glucose transporter isoform in skeletal muscle (28,31). 4 Curiously, GLUT2 expression in GLUT4 null liver is significantly increased. The increase in GLUT2 expression suggests that the liver is capable of increased hepatic glucose uptake. This excess glucose subsequently could be converted to fatty acids or glycogen.
Though GLUT4 null hearts display characteristics of hypertrophy caused by hypertension, they have normal blood pressure (1). The GLUT4 null heart represents a unique model of hypertrophy that may be used to study the consequences of altered substrate utilization in both normal and pathophysiological conditions. Consistent with this notion, GLUT1 expression is increased in GLUT4 null hearts, and serum free fatty acids are reduced, as are peripheral adipose tissue depots. The GLUT4 null heart may in fact be glycolytically primed, which could present an advantage under ischemic conditions. Indeed, preliminary studies demonstrate that GLUT4 null hearts resist loss of function following ischemia/reperfusion. 5 GLUT4 ablation also results in an extreme depletion of fat mass. Gonadal fat pad weights are reduced 10-fold, and fat cells are approximately 50% smaller in size. This reduction in cell size may affect the secretion of the metabolic modifiers mentioned earlier (23). Interestingly, endurance exercise training results in reduced fat mass, smaller fat cell size, cardiac hypertrophy without hypertension, and reduced fed serum free fatty acids (32). The striking similarity between GLUT4 null mice and endurance exercisetrained athletes may suggest that similar adaptive responses are elicited by these two forms of cellular stress.

Characterization of a Novel Glucose Transport Activity in GLUT4 Null Soleus Muscle
The ability of two GLUT4 null muscle types to take up glucose in the presence of maximally stimulating concentrations of insulin was measured in vitro (29 -31). Perhaps as expected, fast twitch EDL muscles of GLUT4 null mice failed to take up more glucose in response to insulin. Surprisingly, a 2-fold increase in insulin-stimulated glucose uptake was noted in female GLUT4 null slow twitch soleus muscles compared with a 3-fold increase in wild type controls. Soleus muscle of GLUT4 null males displayed a 2-fold increase in basal glucose uptake with no further increase following insulin stimulation. The molecular basis for the sexually dimorphic response to GLUT4 ablation in soleus muscle may be linked to the superior insulin sensitivity of female mice. These results indicate that highly oxidative soleus muscle can adapt to ablation of GLUT4 and take up a large amount of glucose, whereas glycolytic EDL muscle cannot. The specificity of glucose uptake was demonstrated by incubating muscles in the presence of 50 M cytochalasin B, a fungal metabolite that inhibits facilitated D-glucose transport (29). Basal and insulin-stimulated glucose uptake in soleus and EDL muscles from GLUT4 null and control mice was reduced by cytochalasin B to the same extent. This result, combined with the failure to detect increased expression of any known GLUT, led to the hypothesis that a novel glucose transport system is responsible for glucose uptake into highly oxidative muscle, which contributes to euglycemia in GLUT4 null mice (1, 28 -31).
Recently glucose uptake was measured in GLUT4 null muscles under normoxic and hypoxic conditions (33). Hypoxia has been shown to stimulate glucose transport in skeletal muscle via a pathway that is independent from that of insulin by recruitment of GLUT4 to the plasma membrane (34). In both soleus and EDL from GLUT4 null mice, hypoxia treatment failed to stimulate glucose uptake to levels above basal normoxic conditions (33). This result proved GLUT4 is essential for hypoxia-induced increase in glucose uptake, and the compensatory glucose transport activity in GLUT4 null soleus does not respond to stimulation by hypoxia in vitro. As hypoxia is a useful model for exercise, the above data suggest that GLUT4 is essential for exercise-stimulated increases in muscle glucose uptake.
Basal and insulin-stimulated insulin receptor tyrosine kinase activity was shown to be normal in muscles of male GLUT4 null mice (31). Furthermore, insulin receptor autophosphorylation was also shown to be unchanged in null muscle. Thus, GLUT4 ablation does not alter the activation of insulin receptor tyrosine kinase activity in skeletal muscle. Glycogen synthase, like GLUT4, is a major downstream target of insulin receptor action (35). Basal glycogen synthesis was increased in GLUT4 null soleus muscle; however, insulin stimulation resulted in a parallel decrease in glucose uptake and glycogen synthesis (31). Although GLUT4 null soleus muscles are able to maintain nearly normal steady state levels of glycogen, acute insulin-stimulated glucose uptake and incorporation into glycogen appear to be GLUT4-dependent. The increased glycogen content of female GLUT4 null soleus muscle was tightly associated with maintenance of ATP and phosphocreatine levels similar to controls in response to hypoxic stress (33). Though GLUT4 null muscles take up significantly less glucose, they maintain normal high energy phosphate stores possibly because of increased utilization of fatty acids (36). The reduced fed serum free fatty acids in GLUT4 null mice are consistent with this hypothesis.

MLT-GLUT4 Null Mice (Muscle Only GLUT4 Expression)
Mice expressing GLUT4 only in fast twitch skeletal muscle were generated to assess the role of muscle GLUT4 in whole body glucose disposal, insulin sensitivity, energy homeostasis, and the complex phenotype of the GLUT4 null model. The muscle-specific MLC-GLUT4 transgene (26) was mated into the genetic background of GLUT4 null mice to generate a population of MLC-GLUT4 null mice (37). MLC-GLUT4 null mice have restored GLUT4 expression in glycolytic muscle such as EDL but not in oxidative muscles like soleus. Fed serum lactate levels of MLC-GLUT4 null mice compared with GLUT4 nulls are restored to normal. Unlike GLUT4 null mice, MLC-GLUT4 null mice do not exhibit fed hyperinsulinemia, suggestive of improved insulin sensitivity. Insulin tolerance tests demonstrate that MLC-GLUT4 nulls clear glucose as well as wild type control mice, unlike GLUT4 null mice, which are significantly insulin-resistant. This result was corroborated by a significant increase in insulin-stimulated glucose transport into EDL muscles. Furthermore, MLC-GLUT4 null soleus muscle retained the compensatory glucose transport activity measured in GLUT4 null soleus. When combined, these findings indicate that expression of the compensatory glucose transporter is not dependent upon hyperinsulinemia.
Like GLUT4 null mice, MLC-GLUT4 null mice weigh approximately 20% less than controls and exhibit significantly decreased fed serum free fatty acid levels (37). Further, MLC-GLUT4 null mice also have significantly reduced inguinal fat pads and adipose cells compared with controls. These data demonstrate that complementation of fast twitch muscle with GLUT4 does not correct defects in adipose tissue mass or lipid metabolism because of GLUT4 ablation. When combined, these data indicate that muscle GLUT4 is a major regulator of whole body glucose metabolism and that defects in glucose metabolism and adipose tissue mass and lipid metabolism in GLUT4 null mice arise independently.

Conclusion
To date GLUT4ϩ/Ϫ mice represent the only gene knockout model where the phenotype of the heterozygote is more severe than that noted in the homozygote (with respect to the diabetes). Additionally, this is the first mouse model that demonstrates that a heterozygous knockout of a single known diabetogene can yield the plethora of pathologies that characterize type II diabetes with increasing age. In contrast to GLUT4ϩ/Ϫ mice, which express normal amounts of GLUT4 in their insulin-responsive tissue until 2 months of age, GLUT4 null mice never express GLUT4 and do not develop a diabetic phenotype. The mechanisms that develop in GLUT4 null mice to help maintain euglycemia may not be apparent in GLUT4ϩ/Ϫ mice because the function of the compensatory glucose transport system may be masked by the remaining functionally dominant GLUT4. Perhaps, endurance exercise training that elicits characteristics seen in GLUT4 null mice (i.e. reduced adipose tissue, cardiac hypertrophy, reduced serum free fatty acids) might lead to expression of the glucose transport system seen in GLUT4 null soleus muscle. This expression could be responsible in part for the therapeutic effects of exercise by improving whole body insulin sensitivity in diabetics (32). Understanding the genetic and molecular basis of the adaptive responses to GLUT4 ablation responsible for the maintenance of euglycemia will undoubtedly lead to novel therapies for diseases of insulin resistance including type II diabetes.