This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Copyright Permission Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Mitanchez, D. Articles by Kahn, A. Search for Related Content PubMed PubMed Citation Articles by Mitanchez, D. Articles by Kahn, A.

Glucose-Stimulated Genes and Prospects of Gene Therapy for Type I Diabetes¹

Institut Cochin de Génétique Moléculaire, Unité 129 de l’INSERM, Centre Hospitalo-Universitaire Cochin, 75014 Paris, France

	Abstract

Top Abstract I. Introduction II. Transcriptional Regulation... III. Cell and Gene... References

 

I. Introduction

II. Transcriptional Regulation of Glucose-Responsive Genes

A. The insulin gene

1. Glucose-signaling pathway in pancreas ß-cells

2. E elements of the insulin gene promoter

3. A elements of the insulin gene promoter

4. Role of the E and A boxes of the insulin gene promoters in glucose-dependent transcriptional regulation

5. Role of the cAMP pathway in glucose responsiveness of the insulin gene promoters

B. Transcriptional regulation of glucose-responsive genes in the liver

1. The general outlook for glucose metabolism in the liver

2. Transcriptional glucose signaling mechanisms in the liver

3. Glucose/carbohydrate response elements

III. Cell and Gene Therapies for IDDM

A. Islet and cell transplantations

1. Transplantation of islets

2. Transplantation of established lines

B. Bioartificial pancreas

C. Transplantation of genetically engineered cells

1. The "ideal" engineered cell for glucose-dependent insulin secretion

2. Engineering established neuroendocrine cells

3. Engineering endogenous hepatocytes

	I. Introduction

Top Abstract I. Introduction II. Transcriptional Regulation... III. Cell and Gene... References

 
GLUCOSE is an important modulator of gene expression in practically all living cells and organisms, procaryotes as well as eucaryotes, yeasts as well as multicellular plants or animals. However, the molecular mechanisms underlying the action of glucose on the transcriptional machinery have remained relatively obscure until recently and are still imperfectly understood. In vertebrates, glucose action can be direct, being mediated by glucose itself or by glucose metabolism, or indirect, secondary to glucose-dependent modifications of hormone secretion, mainly of insulin and glucagon. However, the insulin and glucose pathways, although interdependent, have recently been better defined, and glucose-response elements have been described in various genes specific to the endocrine pancreas, liver, and adipocytes (1, 2, 3).

Insulin-dependent diabetes mellitus (IDDM), or type I diabetes, is a chronic disorder that results from autoimmune destruction of the insulin-producing pancreatic ß-cells. In the United States, the lifetime prevalence of IDDM is about 0.4 %. Thus, with a population of 250 million, approximately 1 million Americans have IDDM (4). Exogenous insulin the-rapy, which is the main treatment, does not permit a glycemic control as precise as that provided by natural secretion from functional islet ß-cells, and acute decompensations and long-term complications are always present. It is therefore appropriate to look at possible alternative therapeutic strategies for IDDM.

Since insulin synthesis and secretion is physiologically regulated by glucose, the identification of the glucose-signaling pathway and transcriptional response elements suggests that it would be possible to replace destroyed endocrine ß-cells in type I diabetes by engineered cells expressing an insulin transgene controlled by glucose-sensitive regulatory regions efficient in these cells. This review will present the state-of-the-art in gene regulation by glucose in the pancreas and liver and will describe how the acquired knowledge in this field could lead to the development of a gene therapy strategy for IDDM.

	II. Transcriptional Regulation of Glucose-Responsive Genes

Top Abstract I. Introduction II. Transcriptional Regulation... III. Cell and Gene... References

 
Many genes implicated in metabolic pathways are transcriptionally activated by high-carbohydrate diets in pancreas, liver, and adipocytes (Table 1). However, studies carried out in normal animals or diabetic animals receiving insulin do not allow the determination of the respective involvement of glucose or of insulin alone or combined in the observed glucose-dependent gene stimulation. Such a differentiation between the glucose and insulin effects requires the use of cultured cells (primary cultures or established lines) whose extracellular hormonal and nutritive environment can be precisely determined. Certain genes listed in Table 1 have not been studied using such systems, so it is not clear whether glucose, insulin, or both are involved in their regulation. In pancreas, the main glucose-responsive gene is the insulin gene. It has been known for a long time that insulin biosynthesis and secretion are sensitive to the level of circulating glucose. In the liver and adipocytes, some glycolytic and lipogenic enzyme genes are transcriptionally induced in response to carbohydrates. This response allows the organism to better utilize limiting carbohydrates in the environment and to store energy.

In this review, we will especially focus on the transcriptional regulation of genes whose mechanisms of glucose responsiveness have been intensively studied: the insulin gene in the endocrine pancreas and the L-type pyruvate kinase (L-PK) gene in the liver. Most of the mechanisms discussed for the L-PK gene are probably relevant to glucose-dependent regulation of other genes in the liver and adipocytes, the most widely studied of which is the Spot14 (S14) gene.

A. The insulin gene
Insulin is one of the most important regulators of glucose homeostasis. It is produced specifically by pancreatic ß-cells. When the glucose concentration rises, insulin is rapidly released from storage granules, and the level of insulin mRNA increases through transcriptional activation and insulin mRNA stabilization.

1. Glucose-signaling pathway in pancreas ß-cells.
Regulation of insulin secretion at physiological glucose levels depends on the pancreatic glucose sensor system. Glucose uptake by these cells is facilitated by the high Michaelis-Menten constant (K_m) glucose transporter Glut2. This step is not rate-limiting, and the cellular glucose concentration rapidly equilibrates with changes in blood glucose concentration. Once in the ß-cells, glucose is phosphorylated to glucose 6-phosphate (G6-P) by the specific hexokinase IV, also termed glucokinase (GK), which exhibits a high K_m for glucose (12 mM). Glucose metabolism in ß-cells generates different signals that are common to many different cell types. The exact nature of the glucose-derived intermediate(s) regulating insulin gene expression is not known, but ATP, acting on an ATP-sensitive K⁺ channel, is thought to play an important role in activating insulin secretion (14, 15). The biosynthesis and secretion of insulin by the islets of Langerhans are not inevitably coupled since these processes can be dissociated under certain conditions (16). Glucose-stimulated insulin release is inhibited in a Ca²⁺-free medium, whereas synthesis is still activated (17). The threshold for glucose-induced activation of insulin synthesis (2.5–3.9 mM) is lower than that for insulin secretion (4.2–5.6 mM) (18, 19). Although glucose is known to stimulate islet insulin mRNA levels over long periods of exposure (2–72 h), blot-hybridization studies using cDNA probes have demonstrated that there is no significant increase in islet insulin mRNA during 1 h exposure to the sugar (20). Over the same time period, incorporation of radiolabeled amino acids into proinsulin is increased 10- to 20-fold (21, 22, 23, 24). This effect does not require the synthesis of new mRNA as indicated by the rapidity of the response and the lack of effect of transcriptional inhibitors such as actinomycin D (22, 25). Thus the initial stimulation of insulin biosynthesis, which occurs within 20 min of exposure to glucose (24, 25), must utilize preexisting mRNA and involve translational regulation.

Glucose metabolism is necessary for its stimulatory effect on insulin gene transcription (26). German et al. (27) have shown that the rat insulin I gene 5'-flanking DNA (rat insulin I promoter) can direct the glucose-regulated transcription of the linked reporter gene chloramphenicol acetyltransferase (CAT) in ß-cells (27), and the glucose stimulation of CAT activity driven by a rat insulin I gene fragment (nt -345 to +1) was inhibited by mannoheptulose (potent inhibitor of GK) in transfected HIT cells (28). Cotransfection of ß-cells with insulin gene reporter constructs, along with cDNA encoding enzymes involved in glucose metabolism, indicated that anaerobic glycolysis could generate a signal (or signals) mediating transcriptional control of insulin gene (29). The nature of this proximal signal (or signals) is unknown.

In most species, including human beings, the insulin gene exists as a single copy. In contrast, rodents possess two nonallelic insulin genes designated insulin I and II, which are almost equally transcribed. Most works have focused on the human and rat insulin gene promoter. While the 5'-flanking sequences of this gene are not as well conserved between species as transcribed portions of the insulin gene, the critical sequences of the promoter are nevertheless quite similar, and their overall functional architecture is retained within mammals, allowing ß-cell-specific transcription of the human insulin gene in transgenic mice (30, 31). The great number of studies on this subject has led to a complex nomenclature, and the names of the different cis-acting DNA elements and trans-acting transcription factors described in various species have not reflected the similar architecture of these promoters. Fortunately, German et al. (32) have recently proposed a nomenclature that uses a single set of designations for all mammalian insulin promoters, thereby making direct comparison simpler (32). This nomenclature, shown in Fig. 1, will be used in this review.

View larger version (19K): [in this window] [in a new window]   Figure 1. New insulin promoter nomenclature. The new names are in boxes, and the previous names are shown below each gene. [Adapted with permission from M. German et al.: Diabetes 44:1002–1004, 1995 (32).]

View larger version (22K): [in this window] [in a new window]   Figure 2. Arrangement of regulatory elements and cognate binding proteins in the promoter of rat insulin I gene. Nucleotides are numbered upstream of the transcription start site. Sequence elements are described below the scheme and the regulatory protein above. [Adapted with permission from K. Docherty and A. R. Clark: FASEB J 8:20–27, 1994 (26).]

4. Role of the E and A boxes of the insulin gene promoters in glucose-dependent transcriptional regulation.
It has long been known that insulin gene transcription responds to glucose. However, the mechanism of the glucose-induced transcriptional response was not well understood. A 50-bp sequence between nt -196 to -247 upstream from the rat insulin I promoter can confer glucose responsiveness to nonresponsive glucose promoters in transiently transfected adult rat and human islets (27). It contains the Flat (or A3/A4) and the Far (or E2) elements and is termed the FF minienhancer (Fig. 2). The Flat element acts synergistically with the E2 or Far element. None of them can function alone, but together they activate transcription of a linked gene (62). The sequence spanning nt -193 to -227 in the rat insulin I promoter is a glucose-sensitive element that binds a ß-cell-specific nuclear factor, the C1 complex whose binding activity is induced by glucose in rat islets. This complex is also able to bind the sequence -206 to -227 of the human insulin gene that is highly homologous to the rat insulin I glucose-sensitive element (63). The rapid change in binding activity of the C1 complex in response to glucose implies modification of proteins that are already present rather than the production of a new protein. MacFarlane et al. (61) demonstrate that IUF1 binds the human insulin gene enhancer in a glucose-responsive manner and that changes in the binding activity of the protein occur as a result of an active modulation of its phosphorylation state in rat islets. The protein IPF1/STF1/IUF1 and C1 factor share the same properties: they are ß-cell specific, their binding activities are modulated by glucose, and they bind to similar sequences in the insulin gene. Thus it appears that these proteins are highly related or identical and are important factor(s) for glucose-dependent regulation of the insulin gene.

These experiments suggest that A boxes are involved in glucose responsiveness. However, it appears that the Flat (A3/A4) element of the rat insulin gene promoter interacts synergistically with the adjacent Far or E2 box. A direct interaction between the ß-cell homeodomain factor lmx1, bound at the Flat element, and the E47 component of IEF1, bound at the Far element, has been demonstrated when the lmx1 and the E47 cDNAs are coexpressed in an insulin-nonproducing fibroblast cell line transfected with the minienhancer (51). STF1-mediated activation of the rat insulin promoter is dependent on the binding of the factor IEF1 on the E1 site and relies on cooperative interactions between the HLH proteins E47 and STF1 (59, 64). More recently, German and Wang (65), using transient transfection of islet cells in primary culture, demonstrated that the E2 box is directly involved in glucose responsiveness. When different mutations in the rat insulin I FF minienhancer are analyzed, the only mutation to remove all response to glucose was located in the CATCTG sequence of the E2 box. In contrast, mutating the binding sites in the contiguous Flat element resulted in a lower glucose responsiveness (3.0-fold vs. 6.0-fold for the wild type). Moreover, nuclear extracts from islets grown in various glucose concentrations demonstrate a glucose-stimulated increase in the protein complex that binds the Far element and contains the ubiquitous transcription factors Pan-1 and Pan-2. Overexpression of intact or partially deleted Pan-1 ablates the Far-directed transcriptional response to glucose, probably because it interferes with the function of glucose-responsive specific b-HLH protein(s) (65). Using deletions and substitution mutations of the proximal human insulin promoter expressed in islet cells in primary cultures, Odagiri et al. (66) mapped a metabolic response element to the E1 box. This isolated E1 element responded to glucose, but inclusion of the A1 or A2 boxes on either side of E1 resulted in dramatic synergistic activation. Sharma and colleagues (67, 68) indicated that glucose-induced transcription of the rat insulin II gene in ß-cell lines is mediated by a sequence containing the two types of cis-acting elements, ICE or E1 box and RIPE3a or A2 box. Thus, several elements within the insulin promoter probably combine to give the full response to glucose.

5. Role of the cAMP pathway in glucose responsiveness of the insulin gene promoters.
As indicated earlier, the glucose-dependent signaling pathways involved in transcriptional control of the insulin genes are unknown. Since cAMP can activate the insulin gene promoters and raise insulin mRNA level, it has been hypothesized that the cAMP pathway might contribute directly to glucose stimulation of insulin gene transcription through the cAMP response element (CRE) present at nt -184 to -197 on the rat insulin I gene and at equivalent positions on the rat insulin II and human insulin genes (69, 70). It has been proposed that a cAMP increase activates protein kinase A (PKA) that phosphorylates CREB, the cAMP response element binding protein, and stimulates CREB binding to its cognate site within the enhancer region (26). However, German and Wang (65) demonstrated that the CRE is not required for glucose response in fetal rat islet, since CRE mutations do not decrease the response to glucose. Therefore, cAMP does not seem to play a direct role in regulating activity of the insulin promoter, but it may modulate the signal supply by other intracellular messengers, involving nutrient metabolism. The potential role of phosphorylation reactions in the effect of glucose on the insulin gene was shown by Docherty and colleagues (61) who demonstrated that IUF1 binding activity in the islets of Langerhans is modulated by glucose in a phosphorylation-dependent manner; however, PKA and protein kinase C do not seem to be involved.

It appears, therefore, that considerable uncertainties persist concerning the mechanisms of the transcriptional response of the insulin gene to glucose. One of the major difficulties is that the results depend on the cell system used (either established insulinoma cell lines or primary cultures of ß-cells or islets of Langerhans). In addition, the effect observed using numerous mutations of subfragments of the insulin gene promoter (e.g., the so-called minienhancer of the rat insulin I gene) are much less apparent using a large regulatory region, indicating that most of the regulatory elements of these promoters have redundant functions. Finally, no report has yet been published on the glucose responsiveness of various insulin gene constructs in transgenic mice. Consequently, it remains to be determined whether the cis-acting elements responsible for tissue-specific and glucose-dependent regulation of the insulin genes identified by ex vivo studies are pertinent in vivo. From the different studies on the "glucose-sensitive elements" of the insulin gene promoters, it can be concluded that glucose responsiveness seems to involve two types of interacting cis-acting elements, termed E and A boxes, that are present in two and three copies, respectively. The E boxes bind heterodimers composed of an ubiquitous and a tissue specific b-HLH protein, while A boxes bind different types of homeodomain proteins, one of them, at least, playing an essential role in pancreas differentiation. Part of the binding activities to both E and A boxes could be regulated by glucose, acting probably by posttranslational mechanisms, i.e., phosphorylation reactions (61). Interactions between several of these elements, especially between E boxes and A boxes, could be required to confer good glucose responsiveness (66, 67, 68).

B. Transcriptional regulation of glucose-responsive genes in the liver
1. The general outlook for glucose metabolism in the liver.
Glucose enters hepatocytes by facilitated diffusion mediated by the Glut2 glucose transporter, whose activity is independent of insulin (71). Upon entering the hepatocytes through the Glut2 transporter, glucose is converted to G6-P by the GK. GK is a member of the hexokinase family. The K_m for glucose is about 12 mM, whereas the K_m for hexokinases I, II, and III ranges from 0.2 to 1.2 mM (72, 73). In hepatocytes, glucose is efficiently converted to G6-P because GK, unlike hexokinases I, II, and III, is not subject to feedback inhibition by the product, G6-P (74). The GK gene is under the control of two alternative liver-specific and pancreas ß-cell-specific promoters (72, 73). While the ß-cell-specific promoter seems to be more or less constitutive in ß-cells, the liver GK gene promoter is activated by insulin and inhibited by glucagon (75, 76, 77). In addition, the enzyme GK is also posttransla-tionally regulated by interaction with a negative regulatory protein whose inhibitory effect on GK is stimulated by fructose 6-phosphate and relieved by fructose 1-phosphate (78, 79). The cDNA for this regulatory protein has recently been cloned (80). G6-P can either be stored as glycogen or metabolized in the pentose phosphate pathway or the Embden-Meyerhof glycolytic pathway (Fig. 3).

View larger version (25K): [in this window] [in a new window]   Figure 3. Glucose metabolism in the liver. Glucose is metabolized to pyruvate and lactate by the glycolysic Embden Meyerhof pathway. Pyruvate is metabolized to acetyl-CoA, which can enter the citric acid cycle for complete oxidation to CO2 and H20, with liberation of free energy as ATP in the process of oxidative phosphorylation. Glucose also takes part in other metabolic pathways, e.g., 1) conversion to its storage polymer, glycogen, by the glycogenic pathway; 2) The pentose phosphate pathway, a source of reducing equivalents (NADPH+) for biosynthesis, e.g., of fatty acids, and also of ribose, which is essential for nucleic acid synthesis. The gluconeogenic pathway utilizes those glycolytic reactions that are reversible, plus four additional reactions that circumvent the irreversible nonequilibrium reactions. The enzymes catalyzing these nonequilibrium reactions are pyruvate carboxylase, phosphoenolpyruvate carboxykinase (PEPCK), fructose 1,6-bisphosphatase (F1, 6-P2ase), and glucose 6-phosphatase (G6-Pase). G6PD, Glucose 6-phosphate dehydrogenase; L-PK, L-type pyruvate kinase; PFK-1, phosphofructokinase-1; PFK-2, fructose 6-phosphate 2-kinase/fructose 2,6-bisphosphatase; GK, glucokinase.

The Embden-Meyerhof pathway is stimulated very early by glucose and insulin via the induction of fructose 2,6-bisphosphate synthesis mediated by insulin-dependent dephosphorylation of the bifunctional fructose 6-phosphate 2-kinase/fructose 2,6-bisphosphatase enzyme (PFK-2). Fructose 2,6-bisphosphate is the major activator of the glycolytic pathway, acting as an allosteric activator of phosphofructokinase-1 (Fig. 3) (74, 84, 85). The pretranslational activation of glycolytic and lipogenic enzymes by glucose and insulin can result from both transcriptional and posttranscriptional events (74, 84, 85, 86, 87). Recently, it has been shown that, in addition to the well recognized role of hormones in mRNA stability, glucose and perhaps other sugars can also regulate the life span of mRNAs independently (88, 89). Similarly, posttranscriptional effects of cAMP have been reported: destabilization of glycolytic enzyme mRNAs (87, 90) and stabilization of a gluconeogenic enzyme mRNA (87, 91).

2. Transcriptional glucose signaling mechanisms in the liver.
a. Respective roles of insulin and glucose.
Glucose, which is a major fuel for mammalian tissues, induces the transcription of several glycolytic and lipogenic genes in hepatocytes and adipocytes (1, 2, 3, 84). The transcriptional effect of glucose can be indirect, being mediated in vivo by hormonal variations, especially increases in insulin and decreases in glucagon secretion. Whereas the transcription of the GK gene, for example, is stimulated by insulin without the aid of glucose (75, 76), the transcriptional activation of most glycolytic and lipogenic genes in hepatocytes requires the presence of both glucose and insulin. In vivo, glucose injection is accompanied by an increase in plasma insulin and a decrease in plasma glucagon concentrations. Insulin administration to animals is obligatorily associated with glucose supplementation for maintaining the blood glucose level. Therefore, studies performed in the whole animal are difficult to interpret due to complex feedback-regulatory mechanisms involving multiple effectors. An excellent way to address this question is by primary culture of hepatocytes since these cells are highly sensitive to both insulin and glucagon and, because of their efficient gluconeogenesis, they survive well in the absence of carbohydrates (75, 76, 79, 92, 93). The availability of various hepatocyte-like cell lines that exhibit different responses to hormones and nutrients is also helpful for investigating the metabolic signaling pathways (94, 95, 96). These culture models are indispensable for investigating the role of glucose and insulin separately and independently from the effects of other hormones; they have allowed us to distinguish clearly between the glucose-independent and glucose-dependent effects of insulin (92).

Transcriptional regulation of the GK gene is mediated by a glucose-independent action of insulin. Although the GK gene is silent in the absence of insulin, the glucose concentration in the culture medium is not a critical factor for insulin-dependent induction of this gene in cultured hepatocytes (75, 76). In vivo in diabetic rats, as well as ex vivo in cultured hepatocytes, the response of the GK gene to insulin is rapid, reaching a high level within 1 h of insulin admin-istration (76). This induction is decreased, but not totally suppressed, by cycloheximide (75, 76), which suggests that insulin action is complex and involves different signaling pathways, one consisting of posttranslational actions (cycloheximide-independent), the other mediated by the activation of a very early insulin-responsive gene (suppressed by cycloheximide). So far, it has been impossible to reproduce the liver-specific hormone-responsive expression of the GK gene liver promoter using cloned constructs in transient expression assays in cultured hepatocytes. Therefore, the cognate hormone response elements of the GK promoter and the actual targets of the insulin action on this gene remain unidentified. In particular, it has been suggested that a fall in the intracellular cAMP level may contribute to induction of GK by relieving cAMP-dependent repression, perhaps through the activation of phosphodiesterase type III (76). In addition, we have demonstrated that the genes encoding the plasma proteins albumin and transferrin are also transcriptionally stimulated by insulin in cultured hepatocytes, regardless of the presence of glucose (92).

Apart from this category of glucose-independent insulin-stimulated genes, most other insulin-sensitive genes so far investigated seem to require the presence of glucose, e.g., the genes for aldolase B (92), PFK-2 (2), fatty acid synthase (97), S14 (98), 5.4 mRNA species (92), and L-PK (92). The role of insulin in the activation of these genes seems mainly to stimulate GK synthesis and thus to permit glucose phosphorylation (94, 99). In hepatocytes, the effect of insulin on the glucose-dependent activation of the L-PK gene can be reproduced by fructose at low concentration (99). This effect is mediated by fructose 1-phosphate, which is the product of fructose phosphorylation by fructokinase (100). Fructose 1-phosphate suppresses the inhibition exerted by a regulatory protein on liver GK (101). In the presence of fructose 6-phosphate, the regulatory protein binds itself to, and inhibits, the liver GK. Fructose 1-phosphate antagonizes this inhibition by inducing dissociation of the GK-regulatory protein complex (100, 101). Furthermore, insulin can be replaced in hepatocytes by transfection of a GK expression vector and, in addition, insulin is not necessary in the glucose-responsive hepatoma cell lines in which GK is replaced by other isoforms of insulin-independent hexokinases (94, 99). Therefore, a major role of insulin in the responsiveness of the L-PK gene to glucose in hepatocytes could be to stimulate the GK-dependent phosphorylation of glucose, i.e., G6-P synthesis.

b. Glycolytic intermediate responsible for glucose-dependent transcriptional activation.
G6-P synthesis is thus the first step of glucose-dependent activation of glucose-responsive genes in liver and hepatocytes (99). As mentioned earlier, G6-P is located at the junction of several metabolic pathways (glycolysis, gluconeogenesis, pentose phosphate pathway, glycogenesis, and glycogenolysis) (Fig. 3). In adipocytes, the glucose analog 2-deoxyglucose (transported into the cell, phosphorylated into 2-deoxyglucose 6-phosphate but not further metabolized in the Embden-Meyerhof pathway) has been shown to stimulate expression of the fatty acid syntase and acetyl-CoA carboxylase (ACC) genes (102). Similarly, 2-deoxyglucose can activate the L-PK promoter in the insulinoma cell line INS-1 (103), but not in hepatocytes or hepatoma cells (94). However, the efficiency of 2-deoxyglucose in mimicking the effect of glucose in some cells does not signify that the observed induction is mediated by 2-deoxyglucose 6-phosphate itself. Indeed, although its isomerization into fructose 6-phosphate is impossible, 2-deoxyglucose 6-phosphate is partly metabolized further into various compounds (104, 105). It has been reported that in some cells (e.g., granulocytes, monocytes, and macrophages), 2-deoxyglucose 6-phosphate can enter the pentose phosphate pathway and be metabolized into a decarboxylated intermediate, most likely a pentose phosphate (104). Therefore, if the 2-deoxyglucose-dependent induction of glucose-responsive genes in adipocytes and INS-1 cells rules out the involvement of the Embden-Meyerhof pathway, it does not rule out the involvement of intermediates arising from 2-deoxyglucose 6-phosphate, especially through the pentose phosphate pathway. The sugar alcohol xylitol is metabolized to xylulose 5-phosphate, an intermediate of the nonoxidative branch of the pentose phosphate pathway (106). In transiently transfected glucose-responsive mhAT3F hepatoma cells, xylitol at low concentration (0.5 mM) induces the expression of a CAT reporter gene controlled by a 183-bp promoter fragment of the L-PK gene (see below), at the same level as 20 mM glucose, while it does not affect intracellular concentration of G6-P significantly (107). The effects of xylitol and glucose on the induction of L-PK gene expression are not additive since 20 mM glucose plus 5 mM xylitol induce the expression of the L-PK/CAT construct to the same extent as 20 mM glucose alone (107). In hepatocytes in primary culture, 5 mM xylitol induces accumulation of the L-PK mRNA even in the absence of insulin (107). Xylitol is oxidized to D-xylulose by L-iditol dehydrogenase and then phosphorylated to xylulose 5-phosphate by a specific xylulokinase (106). Therefore, the xylitol effect in hepatocytes should not depend on the insulin-dependent GK gene activation. Furthermore, the response to xylitol as well as glucose requires the presence of a functional glucose response element (GlRE) (107). It can be assumed from these results that glucose induces the expression of the L-PK gene through the nonoxidative branch of the pentose phosphate pathway. The effect of xylitol at low concentrations suggests that the glucose signal to the transcriptional machinery could be mediated by xylulose 5-phosphate. Recently, Nishimura and colleagues (108, 109) demonstrated the stimulation by xylulose 5-phosphate of a protein phosphatase 2A reacting with PFK-2 . Activation of this protein phosphatase 2A requires at least 10 µM xylulose 5-phosphate, and its activation curve is highly sigmoidal. Xylulose 5-phosphate appears to be a specific activator of the protein phosphatase because none of the other phosphorylated sugars tested, including G6-P or fructose 6-phosphate, was effective (108). Therefore, it could be suggested that glucose induces expression of the L-PK gene through the nonoxidative branch of the pentose phosphate pathway by increasing the concentration of xylulose 5-phosphate. Xylulose 5-phosphate could then activate a protein phosphatase or a kinase(s) and trigger a phosphorylation/dephosphorylation cascade leading to activation of the glucose response complex assembled on the GlRE of the L-PK gene, and most likely of other glucose-responsive genes (Fig. 4).

View larger version (48K): [in this window] [in a new window]   Figure 4. Transcriptional signaling by glucose through the glucose response element of the pyruvate kinase gene in the liver. Glucose enters hepatocytes by facilitated diffusion mediated by the Glut2 transporter. Upon entering the hepatocytes through the Glut2 transporter, glucose is converted to glucose 6-phosphate (G6-P) by the enzyme GK. The liver GK gene promoter is activated by insulin and inhibited by glucagon. The glucose 6-phosphate (G6-P) metabolized through the pentose phosphate pathway could trigger transcriptional activation of the L-type pyruvate kinase (L-PK) gene. Xylulose 5-phosphate (xylulose 5-P), an intermediate of the pentose phosphate pathway, stimulates a protein phosphatase 2A active on fructose 6-phosphate 2-kinase/fructose 2,6-bisphosphatase (PFK-2). PFK-2 activation increases the concentration of fructose 2,6-bisphosphate (F2, 6-P2), a potent allosteric activator of phosphofrutokinase-1 (PFK-1). The xylulose 5-P capable of activating a protein phosphatase activity could trigger a phosphorylation/dephosphorylation cascade, modulating activity of the glucose response complex assembled on the glucose-responsive elements to stimulate the induction of the L-PK gene. We cannot exclude the hypothesis that the xylulose 5-P could activate a kinase as it is able to activate a phosphatase to stimulate the induction of the L-PK gene. F1,6-P2, Fructose 1,6-bisphosphate; AC, adenylate cyclase; Gs, GTP-binding protein; F 6-P, fructose 6-phosphate.

c. Role of the glucose transporters in the transcriptional response to glucose in liver cells.
In pancreatic ß-cells, GK has been shown to be the main glucose sensor responsible for regulation of insulin secretion in response to modification of the plasma glucose concentration (29, 110). However, glucose transporters have also been proposed as possible participants in this glucose sensor function, although this remains controversial. In rodents, pancreas ß-cells possess Glut2 as the main glucose transporter. In contrast, Glut1 is the main glucose transporter in human ß-cells (111). The affinity of Glut1 for glucose is higher than that of Glut2, which explains its low K_m (3 mM) for 3-O-methylglucose. Some cultured rodent ß-cells coexpress Glut1 and Glut2, both being induced by glucose (112). Nevertheless, in insulin-secreting established ß-cells in culture, Newgard and colleagues (113) reported that Glut2 expression was indispensable for a proper induction of insulin secretion by glucose and could not be replaced by Glut1 (113). In support of this, expression of an antisense Glut2 transgene in ß-cells of transgenic mice, which reduced Glut2 expression to about 20% of normal, has been found to disturb insulin secretion (114). It has been proposed that Glut2 could generate an intracellular signal, perhaps in specific interaction with GK, that is needed for the response of insulin secretion to glucose (29).

Glut2 is also expressed in hepatocytes, as well as in other cells that release glucose into the bloodstream (enterocytes and proximal tubular cells of the kidney). Compared with the ubiquitous Glut1 transporter, the characteristics of Glut2 are its previously mentioned lower affinity for glucose and its higher efficiency in allowing for glucose efflux (71, 115). Consequently, glucose can enter or leave the cells expressing Glut2 as a function of the respective intra- and extracellular glucose concentrations. The presence of Glut2 in hepatocytes is therefore thought to be important for rapid glucose efflux after gluconeogenesis. We found that in differentiated hepatoma cells that do not synthesize Glut2, transcription of the L-PK gene is independent of the glucose concentration in the culture medium, while hepatoma cells expressing Glut2 regulate induction of the L-PK gene as a function of glucose (115a).

3. Glucose/carbohydrate response elements.
The first GlRE responsible for the glucose-dependent regulation of glycolytic and lipogenic genes was delineated in the promoter of the L-PK gene. The regulatory region responsible for the transcriptional response of this gene to carbohydrates and hormones was first ascribed to a 183-bp minimal promoter fragment (116, 117). In this minimal promoter we had previously identified the important cis-acting DNA elements, namely, in the 3'-5' direction upstream of the TATA box, box L1, a binding site for HNF1; box L2, a binding site for nuclear factor I (NF1); box L3, a binding site for hepatocyte nuclear factor 4 (HNF4); and box L4, a weak in vitro binding site for USF proteins (Fig. 5) (118, 119). Recently, it had been shown that box L3 was also able to bind NF1 (M. Raymondjean, unpublished data, and Ref 119a).

View larger version (14K): [in this window] [in a new window]   Figure 5. Cis-acting elements and cognate binding proteins identified in the L-PK gene promoter.

The GlRE is therefore both a positive glucose response element and a negative CRE. In transiently transfected hepatocytes in primary culture, adding the GlRE to a glucose-unresponsive truncated L-PK promoter results in a clear inhibition of promoter activity in the absence of glucose or in the presence of glucagon or cAMP (120, 123). Therefore, the GlRE could also be considered to be a silencer whose negative action is stimulated in the absence of glucose or by cAMP. This view has been recently confirmed in transgenic mice: the GlRE is able to blunt the action of a strong enhancer on a contiguous L-PK promoter when animals are fasted or treated with glucagon (124).

The L-PK GlRE consists of two imperfect E boxes: CACGGG separated by 5 bp and differing from the MLTF/USF E box (CACGTG) by a single nucleotide. The two L4 E boxes are indispensable for the GlRE activity.

The L-PK GlRE seems functionally closely related to the carbohydrate response element, ChoRE, described in the regulatory region of the S14 gene (98). This element includes between nucleotides -1432 and -1441 a consensus E box, ACCACGTGAG. As in the case of the L-PK GlRE, mutation of the S14 USF-binding site suppresses the response to glucose, and multimerized S14 fragments, spanning nucleotides -1457 to -1428, confer this response on a heterologous promoter. However, in their normal context, both the L-PK GlRE and S14 ChoRE need to cooperate closely with contiguous DNA elements to confer a good response to glucose (120, 123, 125). This element is box L3, which binds HNF4 and NF1 in the L-PK promoter; the same factor NF1 seems to bind close to the ChoRE of the S14 gene (M. Raymondjean, unpublished data, and Ref. 119a). USF-binding sites are not by themselves glucose/carbohydrate response elements: the GlRE/ChoRE activity requires the special architecture of two functional binding sites separated by exactly 5 bp. Any inactivation of one of the MLTF/USF binding sites or modification in the length (but not in the nature) of the 5-bp spacer results in loss of glucose responsiveness (125). In contrast, elements composed of two high-affinity USF binding sites with the correct spacing are efficient glucose/carbohydrate response elements that, in addition, seem to be fully efficient without the help of contiguous binding sites for auxiliary factors (121, 125). However, there is no correlation between the affinity of the GIRE/ChoRE for MLTF/USF and its efficacy as a glucose response element: mutants with very low affinity for these factors remain active in mediating glucose responsiveness (125a).

Another type of glucose-response element has recently been investigated in the promoter of the ACC gene. In cultured preadipocytes, the glucose-responsive region has been assigned to a -340/-249 promoter fragment containing two GC boxes located immediately downstream of two CACGTC E boxes (126).

a. Transcriptional factor binding to the glucose response elements.
The GlRE/ChoRE of the L-PK and S14 genes are different from the insulin promoter E boxes also involved in glucose responsiveness of the insulin genes. Indeed, while these latter bind especially class A and B b-HLH proteins (see above), the former bind proteins of the b-HLH-LZ family, characterized by the existence of a second dimerization motif at the C-terminal end of the molecules (127, 128).

In addition to USF, this family includes the c-myc oncogene product and its partner Max (129), the immunoglobulin enhancer binding proteins TFE3, TFEB, TFEC (130), and FIP, a Fos interacting protein (131), similar to USF2 (132, 133). All of these proteins form homo- or heterodimers, especially with USF subunits (133, 134, 135). These factors, as well as USF, are ubiquitously expressed, and their abundance is not regulated by diet and hormones. This, however, does not preclude their possible involvement in glucose-dependent transcriptional regulation, because we have shown by in vivo footprinting experiments that dietary modifications associated with turning on and off the L-PK gene transcription do not modify the in vivo occupancy of box L4 (S. Lopez, unpublished observation). Similarly, in yeast a constitutive b-HLH/LZ factor termed centromere binding factor 1 has been shown to be involved in transcriptional activation of sulfur metabolism genes (136). It could be, therefore, that glucose responsiveness results from posttranslational modifications of the complex assembled on the L4 element in fasting as well as in refed rats. Such functional regulations of transcription factors by posttranslational modification are well known phenomena, most frequently involving phosphorylation/dephosphorylation reactions (137). Modification of the redox potential has also been implicated in some cases, e.g., for MLTF/USF, whose DNA binding activity and transactivating potential are modified (138). We recently found that while USF factors do not possess sites phosphorylated by cAMP-dependent protein kinase, they can be phosphorylated by mitogen-activated protein kinases and, indeed, were phosphorylated in vivo. However, their transactivation efficiency is not modified by this phosphorylation more than 2-fold, such that the physiological significance of this phenomenon remains uncertain (L. Gourdon, unpublished data). Finally, it could be that b-HLH-LZ proteins bound at the GlRE serve as "docking transcriptional factors" interacting with glucose-regulated proteins. An example of this type of phenomenon is given by the yeast MIG protein that binds upstream of glucose-repressed genes and acts by recruiting the repressor protein Ssn6-Tup 1 (139).

The involvement of USF in the glucose response of the L-PK gene has been checked by transfecting glucose-responsive hepatoma cells with expression vectors for native or mutant USF proteins (140). We found that overexpression of native USF proteins stimulates the L-PK promoter regardless of the presence of glucose, while overexpression of transdominant negative USF proteins, devoid of either the transactivation domain or the DNA-binding domain, inhibits the glucose responsiveness. Results with the latter mutant are especially interesting since this truncated USF protein is still able to specifically dimerize with endogenous USF proteins but not to bind to the GlRE; it therefore results in specific titration of endogenous USF proteins. These data suggest that USF factors, most likely the USF1/USF2 heterodimer (140), are indispensable components of the glucose response complex. Since our in vivo footprinting data show that the GlRE is occupied regardless of the dietary status of the animals, and thus of the transcription rate of the L-PK gene, it could be that USF-dependent transactivation is regulated positively by glucose and negatively by cAMP. This transactivation potential could be blunted in the absence of glucose or in the presence of cAMP, perhaps by association with a negative regulatory USF partner, such as the Id protein. Overexpression of native USF could titrate this inhibitor, therefore activating the L-PK gene in the absence of glucose, whereas overexpression of negative transdominant USF mutants could impair the formation of the glucose response complex. We are currently attempting to isolate USF partners that could be the targets of glucose and cAMP-dependent regulation. Since the negative cAMP action can be mimicked by transfecting hepatocytes with an expression vector for the catalytic subunit of PKA (cAMP-dependent protein kinase) (M. H. Cuif, unpublished data), we assume that this USF partner could be phosphorylated and regulated by PKA. Using homologous recombination we have also created USF-deficient mice. The L-PK and S14 genes were shown to have a delayed transcriptional response to glucose in USF2 -/- mice (140a). Finally, while USF and related proteins have been demonstrated to play a role in the expression of nonregulated genes (132, 133, 134), the ability of the USF-binding site to be involved in a specific response has already been illustrated in the regulatory region of the type I plasminogen inhibitor gene, for which this site cooperates with a CCAAT box to confer responsiveness to transforming growth factor ß (141). As discussed earlier, the normal function of the "glucose response complex" assembled on the GlRE/ChoRE also requires a close interaction with contiguous auxiliary factors, e.g., HNF4 in the L-PK promoter. This factor is expected to confer tissue specificity to the response of the L-PK gene to glucose. Similar cooperation between HNF4 and metabolic response complexes has been demonstrated in the regulatory regions of the tyrosine aminotransferase (142) and phosphoenolpyruvate carboxykinase (143) genes. For the tyrosine aminotransferase gene, it has been established that HNF4 is indispensable for tissue-specific transcriptional activation by the CRE.

Recently, Viollet et al. have shown that HNF4 was a target for cAMP-dependent protein kinase which induced a decreased binding affinity (143a). Therefore, the factors L3 and HNF4 could also be involved in AMP-dependent transcriptional inhibition of the L-PK gene. Unlike the L-PK and S14 promoters, the promoter II of the ACC gene induced by high concentration of glucose in preadipocytes is not bound or activated by the MLTF/USF family transcription factors, even though it has the recognition sequences (126). The study in preadipocytes and in Drosophila Scheider SL2 cells suggest that the effect of glucose is mediated by the ubiquitously expressed transcription factor Sp1, which can bind to the GlRE of the ACC promoter II and transactivate the gene (126). In addition, it was suggested that glucose could act by stimulating protein phosphatase-1-dependent dephosphorylation of Sp1 (144). However, these experiments do not reveal any information pertaining to the unique features that transform Sp1 binding sites into glucose response elements and do not rule out the possibility that the contiguous E boxes are required for glucose responsiveness.

In conclusion, the glucose-signaling pathway, initiated by glucose phosphorylation to G6-P, targets a glucose-response complex assembled on glucose/carbohydrate response elements of several genes active in the liver, containing two palindromic E boxes with a CACGTG core consensus, separated by five bases. USF proteins binding to these GlRE/ChoRE seem to be required for glucose responsiveness, but this binding is not modulated by glucose. Therefore, USF proteins bound to GlRE/ChoRE are likely to interact with glucose-sensitive partners that have not yet been identified. The example of the ACC gene suggests that such partners could also modulate activity of other transcriptional factors bound at the proximity of the E boxes.

	III. Cell and Gene Therapies for IDDM

Top Abstract I. Introduction II. Transcriptional Regulation... III. Cell and Gene... References

 
IDDM is caused by autoimmune destruction of the pancreas islet ß-cells, leading to insulin deficiency. IDDM is mainly treated with injections of insulin at intervals that are based on the number and content of meals and the sensitivity of each patient to the biological effects of the hormone. Currently, patients require two or more daily insulin injections. A recent study has shown that intensive therapy — i.e., three or more daily insulin injections — is the best treatment to delay onset of chronic complications (145). Glycemic control is never as precise as that provided by natural secretion from functional islet ß-cells. The risk of acute decompensations (hypoglycemia and acido-ketosis) is ever present, and long-term complications, including retinopathy, nephropathy, neuropathy, and cardiovascular disease, continue to cause major health risks for diabetic patients (145). Thus, exogenous insulin therapy is not a perfect treatment for IDDM patients. To restore endogenous insulin secretion, pancreas or combined pancreas and kidney transplantations have been attempted. They abolish the need for daily insulin and in some cases dialysis, but they require chronic immunosuppression, have high failure rates, and may not lessen the chronic complications of diabetes (146). Allografts of isolated ß-islets have the same drawbacks. Moreover, donors of pancreatic tissue are limited, and therefore, the treatment of a high percentage of diabetic patients with islet allografts appears to be unfeasible.

Different ways of search try to improve the treatment of IDDM. We will focus on the different strategies that attempt to replace the ß-cell. The first alternative is to use xenogenic pancreatic islets or ß-cell lines. The second is to genetically engineer nonpancreatic cells to make them able to secrete insulin.

A. Islet and cell transplantations
1. Transplantation of islets.
A Swedish group engrafted porcine fetal islets in diabetic patients. In eight patients, islets were injected directly into the portal vein and in two others the injections were made under the perirenal capsule. All patients had received kidney transplants for kidney failure, and the immunosuppressive regimen had included pre- and postoperative immunosuppressive drugs. In spite of this reinforced immunosuppression, one would have expected the porcine fetal islets to be rejected within days, and this occurred in six patients. In the remaining four, however, pig cells survived for up to 14 months, and porcine C-peptide was detected in the urine, although at very low levels. It is not clear why some patients accepted the fetal pig cells while others did not. Moreover, these cells did not produce enough insulin to have an impact on the patients’ diabetes (147). Recently, Korbutt et al. (148) developed a procedure for isolating neonatal porcine islets in large numbers. When grafted in alloxan-diabetic nude mice, these cells resulted in euglycemia within 8 weeks posttransplantation. Examination of the graft 14 weeks after transplantation revealed a cellular insulin content 20- to 30-fold higher than at the time of transplantation. This indicates that neonatal porcine islet cells have the potential for growth and differentiation in vivo (148).

A method for protecting transplanted tissues from immune rejection has recently been reported by Lau et al. (149). They demonstrated that allogenic transplantation of islets of Langherans was facilitated by the cotransplantation of syngenic myoblasts genetically engineered to express the Fas ligand (Fas-L). When Fas-L interacts with the Fas receptor, found on immune cell surfaces, it induces immune cells to kill themselves in a process called apoptosis. This interaction is thought to play a major role in the maintenance of immunological homeostasis and peripheral tolerance. Lau et al. (149) showed that muscle cells genetically engineered to express Fas-L killed tumor cells expressing Fas in culture. They used these myoblasts to wrap islet tissue and transplanted the composite grafts under the kidney capsule of streptozotocin-induced diabetic mice. This resulted in uniform prolongation of islet allograft survival compared with controls, presumably by killing T cells before they reach the islet cells. Transplanting 2.10⁶ muscle cells lengthened average survival of the islet cell grafts from 10 to 84 days. In about 25% of the mice that received more than 10⁶ myoblasts, blood glucose concentrations stabilized at euglycemic values 10 days after transplantation (149). The downside is that more than half of the mice receiving the highest dose lost their grafts within 80 days, in part because the muscle cells stopped expressing Fas-L. Thus, further investigations are needed to ensure long-term Fas-L expression, which will then permit clinical trials in humans.

2. Transplantation of established lines.
Insulinoma ß-cell-type lines represent an apparently logical alternative to the use of islets. These established lines can be grown in practically unlimited quantity and are therefore much more readily available. The first ß-cell lines were derived from x-ray- or virus-induced tumors (150, 151, 152). Transgenic technology has now been applied to this area. The targeted expression of the SV40 large T antigen in the ß-cells of transgenic mice results in insulinomas that have been used to derive cell lines (153, 154). Such cell lines have attenuated or absent glucose-stimulated insulin secretory responses and, when present, the glucose effect is generally maximal at subphysiological concentrations of sugar. The loss of the response to glucose seems to be exacerbated by time in culture. In normal ß-cells, a specific glucose-sensing apparatus stimulates insulin secretion in response to changes in circulating glucose concentration. As discussed before, it consists mainly of the glucose phosphorylating enzyme GK, and perhaps of the glucose transporter Glut2, both of which have a K_m for glucose in the upper range of physiological glucose concentrations (reviewed in 110 . The loss of glucose response in cultured ß-cell lines parallels alterations in the glucose-sensing apparatus (155, 156). Transplantation of these cells in humans would evidently require previous microencapsulation (see below). An additional hurdle with the transplantation of such immortalized cells in humans would lie in the potential risk of cancer development in the event that the microencapsulation device leaked, although nonprotected cells would be expected to be quickly rejected by the immune system.

B. Bioartificial pancreas
The use of bioartificial pancreas devices capable of excluding immune lymphocytes and immunoglobulins has emerged as a potential alternative to insulin therapy and pancreas or islet allograft. The immune exclusion is achieved by separating islet grafts from the recipient by semipermeable membranes that allow only small molecules, such as glucose, insulin, nutrients, and metabolites, to pass through. Two types of bioartificial pancreas are being investigated: the vascularized ones, which are implanted in the blood vessels; and the nonvascularized ones, which are implanted intraperitoneally or at subcutaneous sites. In a pancreatectomy-induced diabetes model in dogs, Sullivan and colleagues (157, 158) reported that two perfusion devices containing allogenic islets controlled severe diabetes with minimum use of exogenous insulin for up to 1 yr. No immunosuppression was needed to maintain viability of the transplanted islets. Recently, the same results have been reported using xenogenic porcine islets in dogs (159). Implantation of encapsulated rat islets intraperitoneally or at a subcutaneous site in chemically induced diabetic mice resulted in sustained euglycemia for more than 60 days (160). Peritoneal immunoisolated allografts in pancreatectomized diabetic dogs resulted in good control of glucose level in fasted animals, in the absence of any immunosuppression. Treated animals did not need exogenous insulinotherapy, but glycemic control in response to a meal or an intravenous glucose tolerance test remained abnormal (161). These persistent disorders could be ascribed to features of the encapsulation material, and further testing of biocompatible materials is needed to improve the kinetics of insulin secretion and immune protection.

C. Transplantation of genetically engineered cells
Since different problems persist with transplantation of islets of Langerhans or of ß-cells, whether freshly isolated or from established lines, an appealing alternative strategy would be to engineer cells to make them able to secrete insulin as a function of plasma glucose concentration.

1. The "ideal" engineered cell for glucose-dependent insulin secretion.
Ideal engineered cells suitable for transplantation in diabetic patients should possess some features of the endocrine ß-cells allowing them to respond to physiological glucose concentrations: 1) a glucose-sensing apparatus, i.e., the Glut2 transporter and GK; 2) low expression of high-affinity hexokinase isoforms; 3) efficient processing of proinsulin into insulin, which is at least 10-fold more active than the nonprocessed prohormone (162, 163); and 4) efficient regulation of insulin secretion in response to glucose and to other agents. In normal ß-cells, proinsulin conversion into the mature hormone occurs in the secretory granules and is catalyzed by prohormone convertases PC2 and PC3 (164, 165). The hormone is then liberated by granule exocytosis triggered by cell membrane depolarization (166, 167, 168).

Unfortunately, no ß-cell surrogate endowed with all these characteristics exists. However, various strategies aiming at engineering potentially suitable cells have been explored.

2. Engineering established neuroendocrine cells.
More than a decade ago, Moore et al. (169) presented results obtained with a nonislet neuroendocrine cell line derived from ACTH-secreting cells, termed AtT20 line, which was engineered by stable transfection with a human proinsulin cDNA controlled by a viral promoter. AtT20 cells share some features with the islet ß-cells. They contain secretory granules and express the proconvertases PC2 and PC3 (170, 171). Therefore, the AtT20-ins cells released mature insulin polypeptide. Insulin secretion was stimulated by cAMP-related secretagogues and mediated by Ca²⁺ influx through membrane Ca²⁺ channels, a mechanism similar to that mediating the glucose effect in islet ß-cells (172). The GK gene was expressed in AtT20-ins cells under the control of its pancreas-specific promoter. However, as the cells lack the Glut2 transporter, they did not exhibit glucose-stimulated insulin secretion (173, 174, 175). Nevertheless, when stably transfected with a Glut2 expression vector, these cells exhibited glucose-stimulated insulin release, glucose-dependent potentiation of nonglucose secretagogs, and a glucose-sensitive increase in insulin content, all of which are features of normal islet ß-cells. Maximal insulin release occurred at submillimolar concentrations of glucose because the predominant glucose phosphorylating enzyme is not GK but hexokinase, whose affinity for glucose is in the 10 µM range (173). Therefore, as discussed earlier, it appears that normal glucose sensing requires not only Glut2 expression, but also a ratio of GK/hexokinase activities similar to that in the normal ß-cell (111, 113). The importance of this balance between the two glucose phosphorylating activities is well illustrated in insulinoma bTC-7 cells established by Efrat et al. (156) from transgenic mice expressing the SV40 large T antigen in ß-cells. At the initial passages in culture, these cells exhibited a glucose-stimulated insulin secretion analogous to that of islet ß-cells, synthesized Glut2, and had a GK/hexokinase activity ratio similar to that of normal ß-cells. With time in culture, glucose-stimulated insulin release became maximal at low, subphysiological glucose concentrations, in parallel with a 6-fold increase in hexokinase expression (156); in contrast, Glut2 synthesis remained unchanged. Therefore, overexpression of hexokinase appears to be especially deleterious for a physiological response of insulin secretion to glucose. This deleterious effect does not seem to be shared by GK overexpression. Indeed, when human GK is overexpressed in isolated islets infected with an adenoviral GK recombinant vector, the effect on glucose metabolism and insulin release is minimal (176). In contrast, overexpression of yeast hexokinase in ß-cells of transgenic mice (177) and of rat hexokinase I in adenovirus-infected rat islets induces an elevation in basal insulin release (176, 178). To obtain the required optimal ratio of GK to hexokinase in cultured neuroendocrine cells would require both enhancement of GK expression and reduction of hexokinase expression, e.g., by stable transfection of a construct for hexokinase antisense mRNA or knockout of the endogenous hexokinase gene by homologous recombination (174). An additional problem is that AtT20 cells produce the ACTH hormone and possibly other peptides derived from the POMC precursor. Before using these cells, it would be necessary to eliminate this secretion since it is stimulated by glucose in AtT-20-ins cells transfected with a Glut2 expression vector (173). Moreover, the constitutive promoter used in this study for directing expression of the proinsulin gene, i.e., the early promoter/enhancer of the human cytomegalovirus, has been reported not to allow for a prolonged expression in vivo (179, 180). Finally, use of these mouse cancerous cells in humans would raise the immunological and safety problems discussed earlier.

3. Engineering endogenous hepatocytes.
All these potential hurdles to the use of established cell lines, genetically engineered or not, for therapy of IDDM have prompted various groups, in particular our own, to look at the possibility of engineering endogenous cells to transform them into suitable bio-pumps delivering insulin as a function of blood glucose levels. Hepatocytes appear to be especially appealing target cells for somatic gene therapy of IDDM. They are the only cells to possess the same glucose-sensing apparatus as the ß-cell, i.e., Glut2 and GK. Moreover, many liver-specific genes are controlled at physiological glucose concentrations. Finally, the location of the liver is, in theory, privileged for controlling glucose homeostasis, as the portal vein carries glucose and other products absorbed after feeding. However, hepatocytes do not possess secretory granules whose exocytosis is regulated by glucose: no storage compartment for secretory protein is present, and insulin synthesized by these cells will be released via the constitutive pathway (181). Thus, the early insulin secretion that normally occurs when blood glucose levels rise could not be reproduced. However, engineering these cells for gene therapy could provide the basal secretion of insulin needed in IDDM to allow control of metabolism that is required for preventing the complications of diabetes (145). Finally, the liver does not contain the specific proconvertases PC2 and PC3 needed for the cleavage of C-peptide and is therefore unable to completely process the proinsulin molecule (181).

a. Transgenic mice expressing proinsulin gene in the liver.
The possibility of using the liver as a target organ for gene therapy of diabetes has been explored in vivo. Valera et al. (182) obtained transgenic mice expressing the human proinsulin gene under the control of the regulatory regions of the phosphoenolpyruvate carboxykinase (PEPCK) gene. The PEPCK gene encodes a gluconeogenic enzyme. Accordingly, it is transcriptionally activated by fasting and cAMP and inhibited by insulin in the liver (183, 184). These transgenic animals expressing the human insulin gene in the liver were healthy and euglycemic. Proinsulin produced by the hepatocytes was biologically active and could increase GK and glycogen synthase activities, which are highly sensitive to insulin. In the streptozotocin-treated transgenic mice, the insulin mRNA level increased, associated with a parallel increase in human C-peptide in serum and normalization of the expression of insulin/glucose-responsive endogenous genes and of glucose metabolism parameters in the liver, i.e., glycogen and G6-P. The decrease of blood glucose levels in transgenic mice was up to 40% compared with the streptozotocin-treated control mice (184). As indicated above, the PEPCK/insulin chimeric gene is transcriptionally activated by cAMP and glucagon through the CRE of the PEPCK promoter and two CREs in the first intron of the insulin gene (185). During the acute phase of diabetes, -cells of pancreas islets secrete large amounts of glucagon as a response to insulin deprivation, and the glucagon/insulin ratio very actively stimulates the expression of the chimeric gene. However, in the long term, increased insulin production would be expected to block the expression of the transgene since insulin inhibits transcription of the PEPCK gene. Another putative disadvantage of the PEPCK regulatory region for directing proinsulin gene expression is that, in the event of hypoglycemia, the induced glucagon secretion would stimulate transcription of the PEPCK/proinsulin gene, leading to aggravation of hypoglycemia.

b. Improving transgene transcriptional control and prohormone processing in hepatocytes.
In spite of these potential difficulties, this work demonstrates that proinsulin can be produced by hepatocytes and partially compensate for the metabolic disorders secondary to islet ß-cell destruction. However, the weak activity of proinsulin as compared with insulin and the need for a correct regulatory system allowing for activation of insulin production by glucose and inhibition under hypoglycemic/hyperglucagonemic conditions suggest how results obtained with the PEPCK proinsulin transgene might be improved. First, the biological activity of insulin could be enhanced if proinsulin was processed in the hepatocytes. Proteins are processed in eucaryotic cells by either the constitutive or the regulated pathway (186). As already indicated, insulin is normally processed into the mature active A and B chain complex in the secretory vesicles of pancreatic ß-cells that contain the required processing enzymes, proconvertases PC2 and PC3. During processing, the C-peptide that resides between the B and A peptides in proinsulin is excised by enzymes that make two cleavages: one at the B-C junction (an Arg-Arg dibasic site) and one at the C-A junction (a Lys-Arg dibasic site). In cells such as hepatocytes, with only a constitutive pathway, there are no specialized regulated secretory vesicles, and thus proinsulin cannot be efficiently processed. However, partial proinsulin processing can occur in hepatocytes, especially with the rat prohormone; it is catalyzed by a ubiquitous convertase termed furin (181). Furin recognizes the consensus cleavage site Arg-X-Lys-Arg, which is absent in human proinsulin and exists in one copy in rat proinsulin, and consequently proinsulin cannot be processed efficiently by furin. Using site-directed mutagenesis, Gorman and colleagues (187) introduced furin-consensus cleavage sequences into human proinsulin cDNA at both the B-C and C-A junctions (Fig. 6). When transfected into transformed human kidney cells, the variant human proinsulin cDNA was processed into active and mature human insulin (187). Receptor binding and autophosphorylation assays demonstrated that this insulin variant binds and activates the insulin receptor similarly to native insulin. Since furin is well expressed in hepatocytes, this mutant proinsulin is also expected to be efficiently processed in the liver.

View larger version (16K): [in this window] [in a new window]   Figure 6. Processing of proinsulin in endocrine ß-cells and hepatocytes. During proinsulin processing in ß-cells, the C-peptide between the B and A peptides in proinsulin is excised by proconvertases PC2 and PC3. This processing could be reproduced in the hepatocytes by introducing the consensus cleavage site Arg-X-Lys-Arg of the furin.

To test the validity of this strategy, we have created several lines of transgenic mice expressing the human proinsulin cDNA or the mutant construct with the consensus furin cleavage sites under the control of 3.2 kb of 5'-regulating sequences of the L-PK gene. In both cases, mice expressed insulin mRNA in the liver, kidney, and gut. The abundance of this mRNA was regulated as expected by diet and hormones. Human immunoreactive C-peptide could be detected in the serum of the transgenic mice and its level increased under a carbohydrate-rich diet (our unpublished data). However, after streptozotocin intoxication, we did not observe any improvement in transgenic mice expressing the wild type proinsulin or the modified proinsulin construct. We observed that the level of human C-peptide in the serum was very low (D. Mitanchez, unpublished data). This problem was expected to be encountered with the L-PK regulatory region used to direct the proinsulin transgene for two reasons. First, the hyperglucagonemia characterizing decompensated diabetes is likely to block expression of the transgene. Second, activation of the L-PK promoter by glucose requires its phosphorylation to G6-P by GK, which is itself insulin-dependent. Therefore, the response time of the transgene to glucose is expected to be slowed down by the need for prior GK gene activation. To overcome these potential drawbacks of the L-PK system, we are currently investigating new transgenic mice. First, we have added to the L-PK regulatory sequence an heterologous enhancer, from the aldolase B gene (188, 189). It is expected to enhance the level of the proinsulin gene, even in streptozotocin-treated mice. Second, we have created transgenic mice constitutively expressing a GK transgene. In the future, it could be possible to design a composite transgene allowing for constitutive expression of GK and glucose-dependent high level expression of proinsulin to overcome the aforementioned drawbacks of the L-PK promoter. In this connection, the group of Valera and Bosch (190) recently showed that streptozotocin-treated transgenic mice expressing the GK gene under the control of the hepatic PEPCK gene promoter normalized endogenous L-PK gene expression and blood glucose levels. We also plan to mutate the positive glucose response element of the L-PK promoter, which also behaves as a negative CRE (Ref. 120 and this rev