FIELD OF THE INVENTION This invention relates to the field of gene therapy and to a method of utilizing normal non-islet cells transfected with a proinsulin gene inducibly expressed in such cells in the presence of glucose. The proinsulin synthesized in the cells is further processed into mature insulin.

BACKGROUND OF THE INVENTION Insulin-dependent diabetes melitis (IDDM or Type I diabetes) occurs when an autoimmune response destroys the beta cells of the islets of Langerhans, resulting in cessation of insulin production. For many years, and indeed for many patients even at present, the only recourse for treating this fatal condition is the periodic administration of injectable insulin of animal, or more recently, of recombinant human origin.

While the administration of exogenous insulin is life- saving over the long term, severe side effects, such as circulatory disturbances resulting in blindness, gangrene, and heart attack are common. The doses of insulin injected into the diabetic patient are only approximate, even when careful dietary controls are implemented. These continual imbalances in blood glucose resulting from deviations from optimal levels of insulin are thought to cause, or contribute to, the observed side effects.

There have been many attempts to minimize the side effects of insulin therapy. Home blood and urine glucose test kits help the diabetic monitor blood sugar levels. Some researchers have proposed pump devices

which meter insulin continuously and in more precise doses. One interesting such device is disclosed in U.S. Patent No. 5,364,838 which delivers insulin in aerosol form. The hormone is absorbed through the lung, thereby avoiding the more invasive injection route of administration.

In the field of diabetes there is a general recognition that replacement of beta cellular function would be a superior therapy to insulin administration, because the natural cell would secrete insulin in response to glucose levels in the microenvironment.

This fine-tuned control would thereby eliminate the deleterious side effects of insulin administration.

Indeed, it has been shown in the relatively small group of patients who have successfully received pancreatic transplants that remission of side effects or cessation of progressive tissue damage results. Unfortunately, pancreatic transplant is available only to a few diabetics, compared to the numbers of afflicted persons. In addition, transplantation is associated with significant toxicity due to immunosuppressive therapy.

There have been many alternatives proposed for providing the benefits of beta cell replacement without involving organ transplantation. These alternatives involve replacement of beta cell function with actual beta cells or other insulin secreting pancreas-derived cell lines, as discussed in Lacy, et al., Ann. Rev.

Med., 37:33 (1986). Since introduction of exogenous cells into the body is perceived by the immune system as any other allograft, it is necessary to isolate the cells from contact with immunoactive cells and substances. In particular, donor islet cells must be protected from T-cells and macrophages mediating cytolytic processes. One approach is physical immunoisolation, either by microencapsulation or by a microporous chamber. The challenge in these technologies is to overcome the natural process of

foreign body rejection resulting in walling off the implant by a dense layer of fibroblasts. The key to a successful implant is to provide both a biocompatible surface and sufficient porosity to permit some degree of microvascularization. For a general review of these approaches, see Brauker, et al., J. Biomed. Mat. Res., 29: No. 12, 1995. To date, immunoisolation still poses significant problems including loss of viability and function of captive cell populations, foreign body rejection, fissure and inflammation, and immune reactions.

Another approach is to engineer beta cell function into cells derived from autologous tissue or artificially constructed cell lines. U.S. Patent No.

5,427,940 discloses an artificial beta cell produced by engineering endocrine cells of the At-T-20 ACTH secreting cells. A stably transfected cell At-T-20 ins is obtained by introducing cDNA encoding human insulin and the glucose transporter gene GLUT-2 driven by the constitutive CMV promoter. The cell line already expresses the correct isoform of glucokinase required for glucose responsive expression of the insulin gene.

This cell line is responsive to glucose, but is regulated at a level of secretagogue below physiological range. Hence, while the system is of interest, it is not of clinical significance because an animal into which these cells are introduced would be chronically hypoglycemic. Another disadvantage is that the cells, being derived from a heterologous source, bear their distinctive foreign antigens, and must be used in immunoisolation. A further disadvantage is that At-T-20 is a transformed cell liner with potential for unlimited growth.

U.S. Patent No. 5,534,404 discloses another approach to obtaining a correctly secretagogue regulated cell line. Starting with beta-TC-6 cells, subpopulations of cells are selected in an initial stage by a cell sorter capable of recognizing cells

having an increased internal concentration of calcium ion, associated with insulin expression (Ca++ activated fluorescence). After successive passages, cell populations are further selected which respond to glucose in the physiological 4 to l0mM range in a typical sigmoidal curve. For therapeutic use, the cells were encapsulated in alginate bounded by a PAN/PVC permselective hollow fiber membrane according to the method of Dionne (U.S. Patent application No.

PCT/US92/03327).

Valera et al., FASEB Journal, 8: 440 (1994) describes transgenic mouse hepatocytes expressing insulin under control of the PEPCK promoter driven by P-enolpyruvate. The PEPCK promoter is sensitive to the glucagon/insulin ratio and is activated in elevated glucose states. The PECK/insulin chimeric gene was introduced into fertilized mouse eggs. Under conditions of severe islet suppression by streptozotocin, the production and secretion of intact insulin by the liver compensated for loss of islet function.

The strategy of gene therapy for treatment of diabetes is complicated by the complexity of insulin regulation and the structure of the protein itself.

The responsive release of insulin from the beta islet cells is a complex event involving migration of preprocessed protein from cytoplasm to the Golgi apparatus where secretory granules bud off and travel to and fuse with the plasma membrane prior to release.

The initial protein product is preinsulin having an N- terminal signal sequence, which is cleaved during transport to the rough endoplasmic reticulum.

Thereafter the resulting proinsulin is further processed to insulin by removal of the C-peptide joining the two polypeptides of the mature molecule, the A and B chains. In engineering the production of insulin in a host cell, it is impossible to obtain a functional insulin by merely providing the A and B

polypeptides. Synthesis of an intact molecule is necessary for proper folding, and only after the correct conformation is obtained, can the C-peptide be snipped out. Thus, any engineered cell expressing mature functional insulin must have the Kex2 enzyme machinery, including the PC1/PC3 and PC2 endopeptidases, or a functional substitute thereof, as suggested by Newgard, Biotechnology, 10: 1112 (1992).

The control of insulin production and release is further complicated by the regulation of glycolytic flux. It is believed that two proteins are used by the beta cell to sense changes in glucose levels: Glut-2 a specific facilitated diffusion type glucose transporter, and a particular glucose phosphorylating enzyme, glucokinase IV. Both enzymes have a higher Km and Vmax than the other enzymes in their related families. Both also have high affinities for glucose that result in large shifts in activity over the physiological range of glucose concentration. While reduction in GLUT-2 results in depression in insulin production, loss of glucokinase abruptly halts insulin production, and identifies glucose phosphorylation as the true rate limiting step. Transformation of cells with expressible genes for these enzymes appears to restore glucose responsive regulatory characteristics to insulin production, but not infrequently outside the physiological range of control. The experiences many researchers have had underscores the problems inherent in the complexity of the control of insulin production through manipulation of the metabolic utilization of glucose. There is thus a need in the field of diabetes for a new model of insulin regulation and beta cell replacement.

SUMMARY OF THE INVENTION Control of insulin production in synthetic beta cells may be accomplished by alternative regulatory

pathways than through attempted restoration of natural control over a transformed beta cell expression system.

While the actual release of insulin in normal beta cells is modulated through metabolic intermediates, as yet poorly understood, an alternative control is at the level of transcription of the mRNA encoding the proinsulin precursor. It is thus an object of the present invention to provide a control system for expression of the proinsulin gene in a suitable host cell, which is independent of the metabolic effectors and intermediates involved in normal regulation.

It is a further object to provide a replacement beta cell autologous with the patient's own cells to avoid a requirement for immunoisolation of insulin- producing cells. Ideally, a cell population is to be selected which can be engineered to synthesize insulin dependent on regulated gene transcription, without excision and extracorporeal manipulation outside the body. It is a further object to provide a gene therapy utilizing a cell population having intact and normally functioning glucose transporter and phosphorylating system, so that control of insulin production is a function only of transcriptional control. Consistent with this object is provision of an enzyme system capable of generating active insulin or insulin-like analog from proinsulin not subject to feedback intervention of the glycolytic pathway.

In accordance with the present invention, a gene cassette for expression of proinsulin in autologous host cells comprises a nucleotide sequence coding substantially full length proinsulin operably linked to a promoter recognized by an RNA polymerase contained in the host cells, together with a glucose responsive regulatory module having at least two glucose inducible regulatory elements located upstream at the 5' end of the promoter. The cassette is integrated into a vector comprising a replication defective viral genome capable, when infecting a suitable target cell in

vitro, of packaging the vector in a viral particle infective for the autologous host cells. The preferred target host cell is the hepatocyte because liver cells already express GLUT-2 and glucokinase IV sufficiently to generate the appropriate intermediates for glucose regulated transcriptional control of the proinsulin gene in the physiological range.

Hepatocytes also express the endopeptidase furin.

A mutation can be introduced into the reading frame of the proinsulin gene that permits furin cleaving at the appropriate site to obtain substantially complete excision of the C-peptide with appearance of essentially native insulin activity. It is therefore an aspect of the present invention that in the transfection method, a vector is provided in which transcriptionally controlled production of proinsulin is substantially completely converted to the active hormone, which is constitutively secreted into the liver parenchyma in response to elevation in glucose concentration.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic depicting the nucleotide sequence of the glucose regulatory modules C and F respectively.

Figure 2 is a genetic map of the pACCMV.plpA 8.8 kb plasmid containing the cloning sites for the expression cassette for proinsulin, and also several adenoviral genes.

Figure 3a is a genetic map showing the insertion diagram of the expression cassette in relation to various markers on the pACCMV.p.A plasmid. Figure 3b shows the order of genetic elements 5' to 3'.

Figure 4 is a genetic map of the large 40.3 kb pJM17 plasmid used to create the final recombinant vector for transfection.

Figure 5 is a genetic map showing the recombination of vectors pACCMV.plpA and pJM17 to yield the AdC/FAM construct.

Figure 6 is a gel reproduction of a Northern blot of RNA isolated from hepatocytes transfected with the recombinant plasmid vector containing the expression cassette, and cultured in the presence of various levels of glucose.

Figure 7 is a gel reproduction of a Northern blot identical to figure 6, only showing the result of a longer exposure of the CMV control.

Figure 8 is a duplicate experiment of that depicted in Figure 6 only showing the migration position of a control band of mRNA under control of the constitutive CMV promoter.

Figure 9 is a gel reproduction comparing rRNA bonding with mRNA from hepatocytes in the presence of various levels of glucose.

Figure 10 is an autoradiograph resulting from a Northern blot demonstrating the time frame of glucose induced synthesis of insulin mRNA.

Figure 11 is an autoradiograph resulting from an SDS-PAGE gel demonstrating the glucose induced synthesis of insulin in rat hepatocytes from an adenovirus vector under the control a promoter with glucose-inducible regulatory elements.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT In the present invention, replacement beta cells for treatment of diabetes Type I are constructed by transfection of autologous cells, preferably hepatocytes, with a vector expressing proinsulin genetically modified to be cleavable to insulin by an enzyme or enzymes endogenous to the transfected cells.

Initially, a gene cassette is constructed containing the proinsulin gene and control elements suitable for its expression regulated by a secretagogue, preferably glucose.

In the isolation of native proinsulin cDNA, total RNA from normal human islet cells was extracted, and the mRNA fraction was isolated and used as a template in an oligo (dT)15primed reverse transcription reaction. Insulin cDNA (-28bp-443bp) was amplified using sense and antisense oligonucleotides which included restriction sites for KpnI and SalI, respectively. The sequences are shown in Table 1 designated TA423 and TA413, and are listed herein as Seq. I.D. Nos. 1 and 2. Alternatively, cDNA can be isolated according to the methods described in Bell, et al., Nature, 282: 525 (1979) using the primers disclosed therein, but incorporating restriction sites compatible with the selected cloning vehicle. In general, it is desirable to include in the amplified region a portion of the intron flanking the open reading frame of the proinsulin gene. The amplified DNA fragment (-28-443bp) containing the entire coding sequence of human insulin and a portion of the untranslated region from the 5' and 3' ends was subcloned into pBlueScript SK+. pBlueScript SK+ is a 2.96 kb colony-producing phagemid derived by replacing pUC19 polylinker of pBS(+/-) with a synthetic polylinker. This cloning vehicle is described in Shon, et al., Nuc. Acids Res., 16: 7583 (1988).

In a preferred embodiment of the invention, hepatocytes are transfected with a vector having a glucose regulated proinsulin gene. Hepatocytes express an endogenous endopeptidase furin. Although furin is known to cleave proinsulin at its B-C junction, it is very inefficient at cleaving the C-A junction.

Cleavage at both sites is required for excision of the C-peptide required for conversion of proinsulin to active insulin. A single point mutation (T267 to G converts the amino acid sequence KQKR to RQKR producing a modified C-A junction compatible with the specificity of furin. Thus, the proinsulin protein can be processed to insulin utilizing a single endogenous enzyme.

The mutation creating the new C-A junction may be effected by standard methods known in the art. For example, conversion of Lys to Arg can be made in two steps. The sense oligonucleotide (TA403 designates Seq.

I.D. No. 3) including a point mutation corresponding to the desired change in the target region was used with the original insulin antisense oligonucleotide (TA413) to amplify one segment of insulin. Similarly an antisense oligonucleotide (TA404) containing the Lys to Arg mutation was used with the original insulin sense oligonucleotide (TA423) to amplify the second fragment of modified insulin (M1). The two fragments, thus produced, can be purified, and a mixture of them used as template DNA in amplification of C-A modified insulin M1 with oligonucleotide TA423 and TA413. The C-A modified insulin M1 may be subcloned in pBlueScript SK+ at the restriction sites Kpn I and Sal I. In practice, the InsM1 DNA was sequenced and found to be error free. TA413, TA404, and TA423 are shown in Table 1, and are listed herein as Seq I.D. Nos. 2, 4 and 1.

The key aspect of the invention is the control elements which make the transcription of the proinsulin gene responsive to the levels of extracellular glucose.

Since the enzymes GLUT-2 and glucokinase are believed

essential for glucose "sensing", hepatocytes, which produce the enzymes, make a good candidate for a replacement beta cell. Although, the mechanism of "sensing" is not known, Applicants postulated that in addition to GLUT-2 and glucokinase, the rest of the "sensing" machinery would be intact in hepatocytes, including formation of any substances mediating gene expression at the transcriptional level.

To test this hypothesis, Applicants utilized two different glucose inducible regulatory modules. Shih, et al., J. Biol. Chem., 269: 9380 (1994) discloses the consensus glucose-inducible regulatory element (GIRE), which has the sequence CACGTG. Applicants utilized two different GIRE modules in their experiments. The first of these modules, designated sequence C, contains two 21 bp long segments joined head to tail by a 14 bp spacer. The second, and preferred sequence designated F, is based on the module found in fatty acid synthetase. The first 21 bp constitute a perfect match of glucose inducible module but then an 11 bp long segment (10-20 bp of the oligonucleotide), CACGTGGGCGC, is repeated a plurality of times (at least twice, and preferably three to six times), creating a series of glucose inducible elements joined head to tail.

A preferred module is shown in figure 1 as TA418, and listed as Seq. I.D. No. 5, in which CACGTGGGCGC is repeated four additional times, creating five glucose inducible elements (CACGTG separated by five nucleotides) joined head to tail. Thus, the preferred regulatory module for transcriptional responsiveness to glucose is a synthetic oligonucleotide having at least two glucose inducible regulatory elements containing the operative regulatory motif segments CACGTG flanking a nucleotide linker segment, conveniently of the sequence GGCGC. The ends of the double stranded oligonucleotide module are synthesized to include half site restriction sequences to facilitate cloning. For example, for cloning in the preferred defective viral

vector described hereinafter, each sense oligonucleotide starts with a Not I half site on the 5' end, and each antisense oligonucleotide includes an Eco RI half site on the 5' end.

A functional cassette includes the structural proinsulin gene, the glucose regulatory module, and a promoter. The promoter is preferably a relatively strong constitutive promoter normally operative only in the host cell of choice, and responsive to the regulatory module located on its 5' end. In the cassette devised herein, the rat albumin promoter happened to be selected, although many other candidates are known in the art. Using the published sequence, (Heard, et. al., "Determinants of Rat Albumin Promoter Tissue Specificity Analyzed by an Improved Transient Expression System", Mol. Cell. Siol., 7: 2425-2434 (1987)) PCR primers were synthesized containing Eco RI and Kpn I restriction sites, as indicated in Table 1 and designated Seq. I.D. Nos.TA420 (6) and TA421 (7).

Nucleotides 1-184 were thereupon amplified. The amplified rat albumin promoter fragment was purified, and cut with restriction enzymes Kpn I and Eco RI.

After cloning into pBlueScript, the sequence was verified by conventional sequencing techniques.

Preferably the PCR amplification is carried out utilizing the pfu polymerase obtainable from Stratagene, which has a significantly lower error rate than other polymerases.

The cassette comprising, 5' to 3', a glucose regulatory response module, a transcriptional promoter whose level of transcription can be further increased by the glucose response module, and the structural gene for proinsulin genetically modified to be cleavable by a host cell endogenous endopeptidase is spliced together and ligated by conventional techniques. The molecular ends of the polynucleotide cassette preferably have single stranded sequences defining the half restriction site corresponding to complementary

half sites on the vector into which it is to be inserted.

The best available vector is a helper-free replication defective plasmid derived from the adenovirus genome, and described in Newgard, et al., "Glucose-Regulated Insulin Secretion," in Molecular Biology of Diabetes, eds. Draznin, et al., Humana Press: 1992. Figures 2-5 diagram the genetic components and construction of the vector containing the gene cassette. The advantages of this vector include the absence of helper virus, thus preventing propagation of virus and high efficiency of infectivity of host cells. It has the disadvantage of being diluted out of replicating cells, since adenovirus integrates the host cell genome with very low efficiency. Other transducing systems useful in the present invention include certain integrating retroviral systems and another helper-free recombinant adenoviral system disclosed in U.S. Patent No.

5,436,146.

Another advantage of the viral-derived vectors is that delivery to target cells in the intact animal does not require excision of tissue, invitro infection, and reimplantation. Nothing however, would preclude the use of an allogenic source of cells under conditions of immunosuppression. A purified viral stock (2-40 infective units per target cell) may be injected into the hepatic portal vein, with efficient infective rates obtainable as viral particles penetrate the hepatic capillary beds and come into contact with the hepatocytes. In this way, replacement beta sites are generated insitu without disturbing the normal cellular architecture. Further advantages of the present invention will become apparent from the Examples which follow.

As demonstrated in the Examples, hepatocytes transfected with an adenovirus vector containing insulin under the control of glucose-inducible response

elements (GIREs) are subject to control at physiological levels of glucose. Figures 6 and 7 demonstrate that the transfected hepatocytes only initiate insulin mRNA synthesis in response to physiological or supraphysiological levels of glucose.

Figure 10 demonstrates that a vector construct containing 2 GIREs initiates transcription in response to elevated glucose levels in a time-frame comparable to islet cells. Finally, Figure 11 demonstrates the synthesis of insulin in response to physiological levels of glucose in hepatocytes transfected with the proinsulin gene under the control of GIREs.

It will be apparent that any structural gene for which glucose modulated control is desired, may be inserted into the gene cassette by conventional recombinant techniques and expressed in an appropriate host cell. For proteins not requiring further processing, the fucin enzyme of hepatocytes is, of course, superfluous. A number of metabolic diseases for which the present invention has therapeutic value in restoring a glucose response mediated protein function can be identified.

EXAMPLE 1 Generation and Cloning of Insulin Gene Cassettes in the plasmid pACdeltaCMV.

The plasmid pACCMV.pLpA (Fig. 2), used as a vector for generation of replication defective recombinant adenovirus containing genes of interest, was cut to completion with the restriction enzyme Sal I and partially with the enzyme Not I. The 8.3 kb piece of KNA, lacking CMV promoter, was gel purified and used as vector for inserting insulin gene cassettes.

The oligonucleotide pair corresponding to one of the GIREs was mixed with gel purified Eco RI - Kpn I albumin promoter and Kpn I - Sal I InsM1 DNA fragments, the mixture was ligated with the above described

plasmid vector pACACMV. A combination of a Glucose Regulatory Response Modules C or F (Figure 1), albumin promoter and the mutant insulin cDNA, produced a total of two constructs. Integrity of both the constructs was confirmed by sequence analysis. Each of the two constructs was cotransfected with the plasmid pJM17 in the host 293 cell line, as described, to generate recombinant replication-defective adenovirus constructs, namely Ad.CAM1 and Ad.FAM1 (see Figure 5).

EXAMPLE 2 Expression of Insulin in Hepatocvtes at various glucose concentrations.

Rat hepatocytes were prepared by in situ perfusion of 0.5 mg/ml collagenase in supplemented balanced Hank's solution as described (Kreamer et. al. (1986) In Vitro 22, 201-211). The viability of isolated hepatocytes was 90% or better.

Six collagen coated 60 mm plates, each containing 1x106 hepatocytes, were transfected with 5x107 pfu/plate. The transfected hepatocytes were exposed to three concentrations of glucose, 3.3 mM, 5.6 mM and 27.5 mM, in RPMI supplemented with 10% fetal calf serum, 30 g/ml proline, 5 yg/ml insulin, 5 Ug/ml transferrin and 5 yg/ml selenium. After 36 h, one of the two plates at each of the tested glucose concentrations, was used to prepare RNA and the other plated was used to check the viability of the hepatocytes. Hepatocyte viability at all the tested concentrations of glucose were no more than 10% different. Ten microgram of RNA from each sample was electrophoretically resolved on a formaldehyde-2% agarose gel, the RNA transferred to a Nylon membrane, W-crosslinked and hybridized with digoxygenin-labeled insulin cRNA. Detection of the membrane-bound probe was performed by chemiluminescence, results recorded as

multiple exposures on X-ray films for various lengths of time and quantitated by digital image analysis.

Referring to figures 6 and 7, Northern analysis reveals that RNA migrating at the position of polynucleotides of approximately 1.35 kb, corresponding to the predicted size of the proinsulin transcript, is evident only when transfected hepatocytes are cultured in the presence of 27.5 mM glucose. Importantly, no induction of the proinsulin gene over background is indicated at 3.3 or 5.5 mM glucose. Unlike other attempts at constructing an artificial inducible insulin-producing replacement cell, in which induction occurs at subphysiological levels of glucose, the present transfected hepatocytes show a response only in physiological or supraphysiological range. Strong induction is seen at glucose concentrations of greater than 5.5 mM. A strong response is apparent at l0mM.

The gels show that both glucose regulatory modules, as described, are functional to about the same degree, although the AdFAM construct using the F module appears to be somewhat more responsive. As a control, the Ad.CMP-Ins in which the gene of interest (proinsulin or beta-galactosidase) is under control of the constitutive CMV promoter, generates RNA of the distinctive size without regard to glucose concentration and is used to quantitate the amount of insulin mRNA (Figures 6, 7 and 8).

Quantitation by phosphoimaging is summarized in Table 2. The results show only a slight difference in relative expression between 3.3 and 5.6 mM glucose, in contrast to a 3.06 value at 27.5 mM. Figure 9 shows the gel normalizing rRNA and mRNA to arrive at the 3.06 value.

Table 1 Oligo Sequence (5'-3') TA403 GAG GGG TCC CGG CAG AAG CGT GGC A TA404 ACG CTT CTG CCG GGA CCC CTC CAG G Sal 1 TA413 CGG AGT CGA CCA TCT CTC TCG GTG CAG GAG GCG G Eco R1 TA4 2 0 GGA ATT CTC TAG AGG GAT TTA GTT AAA CAA CTT Kpn 1 TA421 GGG GTA CCA GAG GCA GTG GGT TGA CAG GT Kpn 1 TA423 GGG GTA CCA TCA GAA GAG GCC ATC AAG CA

Table 2 Quantitation of Insulin mRNA in Hepatocytes Glucose hlns mRNA 18s RNA Normalized hlns hlns mRNA (mM) mRNA (hlns (Relative* mRNA/18s RNA) Expression) 3.3 81074 56 787 0.91 5.6 51039 33 865 1.00 27.5 201056 40 2645 3.06 *For the sake of comparison, the normalized amount of insulin mRNA expressed at euglycemic level (5.6 mM glucose) is arbitrarily assumed to be one.

EXAMPLE 3 Time frame of glucose induced svnthesis of insulin mRNA Collagen coated 60 mm plates or dishes, each containing 1 X 106 hepatocytes, were transfected with 3.5 X 106 pfu/plate of the test adenovirus. Control plates received either no virus, or virus encoding bacterial B-galactosidase. Hepatocytes were then exposed for 16h to 5.6 mM glucose in RMPI supplemented with 10% fetal calf serum, 30 Hg/ml proline, 5 g/ml insulin, 5 Ug/ml transferrin and 5 yg/ml selenium, at 37 degrees Centigrade. The plates containing transfected cells were then divided into two groups, one group receiving fresh medium containing 5.6 mM glucose, the second group receiving fresh medium with 27.5 mM glucose. From each of these two groups, individual plates were removed after 30 min, lh, 2h, 4h, 8h, and 16h, the medium decanted, and the cells frozen in liquid nitrogen. Total RNA was extracted and analyzed for hIns mRNA by Northern blotting.

The Northern blot in Figure 10 demonstrates that after exposure to 27.5 mM glucose, hIns mRNA was detectable at the first time point of 30 min and increased thereafter in a time-dependent manner. At the normal glucose level (5.6 mM) the signal was much lower. Quantitation of the bands revealed that the upregulation of insulin message observed at 27.5 mM is roughly 10 fold as compared to the 5.6 mM treatment.

These data demonstrate that the vector construct containing two GIREs initiates transcription in response to elevated glucose levels in a time-frame comparable to islet cells. This rapid temporal response to elevated glucose levels confers a level of control of insulin synthesis in addition to control mechanisms arising from glucose detection by the GLUT-2 and glucokinase pathways. These data also demonstrate a precise and correct response to physiological levels of glucose. Insulin mRNA production is not upregulated in the presence of 5.6 mM glucose, the steady state concentration of glucose in the bloodstream. Insulin mRNA synthesis is stimulated by elevated (above 5.6 mM) levels of glucose.

EXAMPLE 4 Glucose induced synthesis of human insulin in rat hepatocytes from an adenovirus vector with human insulin (hIns) cDNA under the control of a chimeric serum albumin promoter containing glucose-inducible regulatory elements (GIREs) Freshly prepared rat hepatocytes were transfected with two different adenovirus constructs containing the M1 mutated insulin gene: AdSAM1 (containing the rat Albumin promoter modified to contain 2 copies of the "S14" glucose-inducible regulatory element) and AdCMVInsM1 (containing the constitutive and highly active CMV promoter) . Hepatocytes transfected with AdCMV. -Gal and untransfected hepatocytes were used as controls. Four plates of hepatocytes were transfected

with each adenovirus preparation; two plates were exposed to the low (3.3 mM) glucose and the other two plates to the high (27.5 mM) glucose concentration.

After 36h, hepatocytes were exposed for 16h to 5.6 mM or 27.5 mM glucose in RMPI supplemented with 2 mg/ml bovine serum albumin with leucine omitted, 30 yg/ml proline, 5 Ug/ml insulin, 5 yg/ml transferrin and 5 yg/ml selenium, at 37 degrees Centigrade.

Following a 6h leucine depletion, a 2 ml aliquot of the low or high glucose containing defined medium was added to appropriate plates. For each adenovirus used, one plate for each glucose concentration received 0.2 mCi3H-leucine (500 Ci/mmole). The remaining plate received the equivalent amount of unlabelled leucine and at the end of all incubations it was used for viability determination.

The leucine incorporation was carried out for 16h, followed by a 4h chase with unlabeled leucine. The culture medium was aspirated, cell debris removed, and the supernatant used for analysis of secreted products.

The cells on each plate were lysed with 0.8 ml solution containing 20 mM Tris-HCL buffer at pH 7.6, 2 mM EDTA, 5 g/ml trypsin inhibitor, 50 yM phenylmethane sulphonyl fluoride (PMSF) and 1% Triton-X100. The lysate was centrifuged at 16,000xg for 10 min in a microcentrifuge, the pellet discarded, and the supernatant solution used for analysis of labeled intracellular products.

For each analysis of secreted insulin, 0.8 ml of culture supernatant was used, and for each analysis of intracellular insulin 0.4 ml of cell lysate supernatant was used. Samples were pre-cleared with Staph. aureus in the absence of specific antibodies, to reduce the non-specific precipitation of labeled proteins during the procedure: SOjil of 10% suspension of formalin- fixed Staphylococcus cells (from Calbiochem) was added to each tube; tubes were kept at room temperature for 30 min with continuous mixing, centrifuged (4 min,

16,000xg) and the supernatant used for further analysis. For immunoprecipitation of insulin and insulin-related products, 2.5jim of polyclonal guinea pig anti-human insulin (from Sigma Chemical Co.) was added to the supernatant solution from the above step, mixed and kept at room temperature for 45 min, followed by addition of Staphylococcus cells with 30 min incubation plus centrifugation exactly as before.

Supernatant was discarded, and the pellet washed 4-5 times with 1 ml of 20 mM Tris-HCl at pH 7.6, 0.15 M NaCl and 0.1k Triton X-100.

The pellet was suspended in 40 yl solution containing 60 mM Tris-HCl at pH 6.8, 1.2% SDS, 2% - mercaptoethanol, heated in a boiling water bath for 4 min, centrifuged, and the supernatant analyzed by polyacrylamide-SDS gel electrophoresis. As an internal control for sample to sample variation, in a second set of tubes 2 jil of rabbit anti-rat albumin polyclonal antiserum was included along with anti-insulin antiserum, enabling co-precipitation of human insulin and the endogenous rat albumin in the subsequent single step. Specificity of the immunoprecipitated material was established by the use of control cells transduced with an unrelated gene, -galactosidase, and untransfected cells. To further confirm the identity of the immunoprecipitated material, in a separate set of tubes unlabeled insulin and rat serum were added to provide competition with labeled insulin and albumin respectively, and tubes processed simultaneously.

The optimum gel system for resolution of insulin B and A chains and rat albumin was found to be an SDS/Tris-Tricine 10-20% linear polyacrylamide gradient based on the description of Schagger and Jagow (Analyt.

Biochem. 166:368-379, 1987). A 15 yl aliquot of SDS- ME-treated immunoprecipitated material from each sample was resolved on the gel along with peptide size markers from BioRad. The gels were fixed, stained,

destained, soaked in "Amplify" solution (Amersham), dried under vacuum, and exposed to X-ray film at -800C.

Our results (Figure 11) show the presence of an anti-insulin antibody-binding band in cell extracts of hepatocytes transduced with both of the insulin cDNA- containing constructs, AdSAM1 (glucose-inducible) and AdCMV.InsM1 (constitutive). As noted above, both constructs contain the M1 insulin cDNA, mutated in the coding region at the C/A junction. AdDAM1 contains the S14 glucose-inducible regulatory element coupled to the rat albumin promoter, and AdCMV.InsM1 contains the cytomegalovirus immediate/early promoter. When AdSAM1 was used, the insulin band appeared only in the hepatocytes exposed to high glucose and not the low glucose. When AdCMV.InsM1 was used, the insulin- positive band was present in approximately equal amounts regardless of the glucose concentration. It should be noted that these results were controlled for possible differences in gel loading and other sources of sample-to-sample variation, including a difference in hepatocyte survival after incubation at 3.3 mM versus 27.5 mM glucose (10-15% lower viability at low glucose), by co-precipitating rat serum albumin and using it as an internal standard.

The size of the insulin positive band was determined to be 7,700 Daltons. This differs from the sizes of mature insulin B and A chains and most likely contains C+P peptides of insulin as a result of incomplete processing proinsulin. The size of rat albumin was determined to be 67,000 Daltons, which compares favorably with the known size. The identities of the observed bands as insulin and albumin were confirmed by the fact that the signals were almost completely ablated when excess unlabeled insulin and normal rat serum were included during the immunoprecipitation. A preliminary determination using digital densitometry revealed that intra-cellular insulin protein expression at high glucose was only

about 20-fold lower when driven by the chimeric albumin promoter employed in AdSAM1 then when driven by the CMV promoter in AdCMV.InsM1. This is highly encouraging since the CMV promoter is the most active known promoter in most in vivo and ex vivo mammalian systems.

It would not be desirable to express insulin at the levels achieved by the CMV promoter.

These data demonstrate the synthesis of insulin in response to physiological levels of glucose in hepatocytes transfected with a vector containing the proinsulin gene under the control of a GIRE. At 3.3 mM glucose, there is no detectable sythesis of insulin; at 27.5 mM glucose, insulin synthesis is readily detectable. The addition of GIREs to the promoter prevents promoter overresponsiveness to glucose so that hypoglycemia is prevented. Therefore, a physiologically regulating and physiologically responsive promoter and gene contruct has been created.

The promoter and gene construct may further be characterized as a physiological acute synthesis promoter, meaning that acute synthesis of insulin and insulin mRNA is triggered by increases in glucose concentration of above the physiological setpoint of 5.6 mM, but not below the setpoint. In contrast, prior art references such as Stewart et al. (J. Mol.

Endocrinol. 11:335-341, 1993) and Wilson et al. (WO 94/26915) have focused solely on the regulation of insulin synthesis and secretion occurring through detection of glucose concentration by the GLUT-2 and glucokinase pathways. This approach has led to hepatocyte cell lines which exhibit both constitutive and acute insulin release. These cell lines synthesize and secrete insulin outside the appropriate physiological range of control. Insulin synthesis and secretion commences at glucose concentrations of less than 5.6 mM in these cell lines, resulting in occurrence or possibility of hypoglycemia. By adding two GIREs to the vector, which provide an additional control mechanism at the transcriptional level, the synthesis of insulin is correctly regulated in the appropriate physiological range. Insulin mRNA and insulin are synthesized at glucose concentrations exceeding 5.6 mM and are not synthesized at glucose concentrations of 5.6 mM. This feature makes vectors containing GIREs uniquely appropriate for treatment of Type 1 diabetes and superior to the prior art vectors.