Gene Therapy with Neurogenin 3 and Betacellulin Reverses Major Metabolic Problems in Insulin-Deficient Diabetic Mice

Abstract

Insulin deficiency in type 1 diabetes leads to disruptions in glucose, lipid, and ketone metabolism with resultant hyperglycemia, hyperlipidemia, and ketonemia. Exogenous insulin and hepatic insulin gene therapy cannot mimic the robust glucose-stimulated insulin secretion (GSIS) from native pancreatic islets. Gene therapy of streptozotocin-diabetic mice with neurogenin 3 (Ngn3) and betacellulin (Btc) leads to the induction of periportal oval cell-derived neo-islets that exhibit GSIS. We hence hypothesized that this gene therapy regimen may lead to a complete correction of the glucose and lipid metabolic abnormalities associated with insulin deficiency; we further hypothesized that the neo-islets formed in response to Ngn3-Btc gene delivery may display an ultrastructure and transcription profile similar to that of pancreatic islets. We injected streptozotocin-diabetic mice with helper-dependent adenoviral vectors carrying Ngn3 and Btc, which restored GSIS and reversed hyperglycemia in these animals. The treatment also normalized hepatic glucose secretion and reversed ketonemia. Furthermore, it restored hepatic glycogen content and reinstated hepatic lipogenesis-related gene transcripts back to nondiabetic levels. By transmission electron microscopy, the neo-islets displayed electron-dense granules that were similar in appearance to those in pancreatic islets. Finally, using RNA obtained by laser capture microdissection of the periportal neo-islets and normal pancreatic islets, we found that the neo-islets and pancreatic islets exhibited a very similar transcription profile on microarray-based transcriptome analysis. Taken together, this indicates that Ngn3-Btc gene therapy corrects the underlying dysregulated glucose and lipid metabolism in insulin-deficient diabetic mice by inducing neo-islets in the liver that are similar to pancreatic islets in structure and gene expression profile.

Neurogenin 3 plus betacellulin gene therapy induces neo-islets in the liver with gene expression profile and ultrastructure closely simulating normal β-cells, reversing major metabolic derangements of diabetes in mice.

Diabetes is characterized by dysregulation of many metabolic pathways resulting from a disruption of insulin signaling due to deficiency in either insulin production (absolute in type 1 and relative in type 2) or action (type 2). Insulin regulates hepatic gluconeogenesis such that during feeding (a high insulin state), hepatic gluconeogenesis is suppressed and during fasting (low insulin state), hepatic gluconeogenesis primarily supports blood glucose levels. During abnormally low insulin states, there is unrestrained hepatic gluconeogenesis, resulting in fasting hyperglycemia, the hallmark of diabetes. However, other pathways are also impacted by the absence of insulin signaling such as the ketogenic and the lipogenic pathways. A correction of hyperglycemia is not necessarily accompanied by a correction of the other metabolic disruptions that contribute to the acute and chronic complications of diabetes.

Exogenous insulin administration is currently the standard therapy for insulin-deficient diabetes, but it produces inconsistent physiological control and is often associated with recurrent hypoglycemia. In an attempt to improve on insulin injections, different laboratories have explored the use of somatic gene therapy to treat diabetes by delivering the insulin gene to the liver of rodents with diabetes. Whereas the treatment has often succeeded in reversing fasting hyperglycemia, it has failed to produce a properly regulated glucose-stimulated insulin secretion (GSIS) response (1,2,3,4). Normal GSIS depends on mainly posttranslational regulation of insulin secretion from pancreatic β-cells (5), a process that cannot be mimicked by the transcriptionally regulated insulin transgenes used in these reports. In addition, the high local concentration of insulin in the hepatocytes in response to the insulin transgene expression has been reported to cause undesirable metabolic consequences such as increased serum nonesterified fatty acids and hypertriglyceridemia (2,6,7). In contrast, islet transplantation, when successfully executed, leads to a good control of hyperglycemia as well as reversal of other metabolic derangements associated with diabetes. The major drawbacks of islet transplantation are: 1) the therapeutic effects are not lasting (8), 2) the shortage of islet donors, and 3) the immunosuppressive therapy is associated with toxic side effects (9). The more recent reports on diabetes gene therapy by delivering islet transcription factors to induce insulin-positive cells in the liver produced mixed results with respect to hyperglycemia reversal (10,11); however, none of these studies examined the efficacy of the maneuver on reversing metabolic derangements other than hyperglycemia.

We previously used adenoviral vectors to deliver the islet lineage-defining transcription factor neurogenin 3 (Ngn3) along with the islet growth factor betacellulin (Btc) to the liver of insulin-deficient diabetic mice. In these animals Ngn3 with Btc reverses hyperglycemia within a week and restores glucose tolerance and plasma insulin levels. We have shown previously that Ngn3 initially induces insulin production from the hepatocytes of these mice and subsequently leads to neo-islet formation from periportal oval cells that secrete insulin in a glucose-responsive manner, stably reversing the hyperglycemia (12).

In this study, we examined the hypothesis that Ngn3-Btc-induced neo-islets are similar to pancreatic islets in their ability to maintain normal whole-body metabolic homeostasis as well as in their global transcriptional profile and ultrastructure. We found that, indeed, Ngn3-Btc gene therapy-induced neo-islets were competent in reversing multiple pathways of dysregulated carbohydrate and lipid metabolism associated with insulin-deficient diabetes. Comparison of the Ngn3-Btc-induced islets and pancreatic islets at the functional level revealed that they share remarkably similar transcriptomes by microarray analysis. Morphological analysis by electron microscopy further showed that the periportal neo-islets induced by the regimen contain secretory vesicles that were morphologically similar to those in pancreatic β-cells.

Materials and Methods

Helper-dependent adenoviral vectors (HDAds)

Mouse Ngn3 cDNA (a gift from Dr. Ming-Jer Tsai; Baylor College of Medicine, Houston, TX) and mouse Btc cDNA were cloned into a pBluescript II KS vector (Stratagene, La Jolla, CA). The fully sequenced cDNAs were then subcloned into pLPBL1 shuttle plasmid with an elongation factor-1α (BOS) promoter and rabbit β-globin polyadenylation signal. We used p-delta28 plasmids as backbone for the HDAds, and these were amplified as described previously (13). Diabetic mice with stable hyperglycemia for at least a week were given a single iv injection of HDAd vectors via the tail vein. To exclude any nonspecific effect related to vector dose, the total vector dose was maintained at 6 × 10¹¹ vp in all treatment groups: (5 × 10¹¹ Viral Particles (vp) Ngn3 + 1 × 10¹¹ vp Btc) and (6 × 10¹¹ vp empty vector).

Animals

We induced diabetes in C57BL6/J mice (from The Jackson Laboratory, Bar Harbor, ME; housed in a pathogen-free vivarium in a 12-h light, 12-h dark cycle with ad libitum chow and water) by ip injections of streptozotocin (STZ, 125 mg/kg · d; Sigma, St. Louis, MO) on 2 consecutive days. Diabetic mice were monitored regularly with body weights and 4-h fasting blood glucoses determined by tail snip and using a One Touch glucometer (LifeScan, Inc., Milpitas, CA). Diabetes was defined as two consecutive blood glucoses 250 mg/dl or greater. HDAds were injected iv via tail vein 2 wk after STZ. Serial plasma glucose levels and body weights were measured. Percent body fat was determined at baseline and 0 and 6 wk after treatment using an EchoMRI whole body composition analyzer (Echo Medical Systems, Houston, TX). Fasting plasma insulin (Mercodia, Uppsala, Sweden), free fatty acids (FFAs; Wako, Indianapolis, IN), triglycerides (Wako), and β-hydroxybutyrate (Sigma) were assayed at baseline and 0 and 6 wk after treatment. A glucose tolerance test was performed, 6 wk after treatment, on mice that were fasted 4 h and injected with 1.5 g/kg of d-glucose ip, and blood collected at 0, 15, 30, 60, and 120 min. Mice were anesthetized and tissues collected and processed for RNA, immunohistochemistry, and immunofluorescence as indicated. Formalin-fixed, paraffin-embedded liver sections were deparaffinized, hydrated, and stained for glycogen using the periodic acid-Schiff staining system (Sigma); counterstained with hematoxylin; and mounted before visualization. We fixed the 8-μm-thick frozen liver sections in 10% buffered formalin for 5 min and, after washing, stained them with freshly prepared, filtered Oil Red O (Sigma) 0.3% in 60% isopropanol for 15 min followed by rinsing in 60% isopropanol and counterstaining with hematoxylin. The staining was visualized using the Axiovert (Zeiss, New York, NY) microscope with Axiovision imaging software (version 4.0). All images within the same experiment were acquired at the same settings and processed identically. Liver glycogen content (glycogen assay kit; BioVision, Mountain View, CA) and triglyceride content (BioVision triglyceride quantification kit) were measured from frozen tissues as per protocol provided with the kits. The values were normalized to the tissue weight.

For experiments involving laser capture microdissection (LCM) (see below), transgenic mouse insulin promoter (mip)-green fluorescent protein (GFP) mice (14), which have GFP driven by the mip facilitating the identification of insulin positive cells were used using the same protocol, as described in this section, for induction of diabetes and for gene therapy.

Quantitative real-time PCR for gene expression analysis

We removed and snap froze livers from anesthetized mice in liquid nitrogen and stored the samples at −80 C. They were homogenized in Trizol reagent (Invitrogen, Carlsbad, CA) and total RNA extracted as per the manufacturer’s instructions. We then subjected the RNA to an on-column purification and deoxyribonuclease I digestion (RNeasy mini; QIAGEN, Valencia, CA) to remove any genomic DNA contamination. The eluted RNA was quantitated and stored at −80 C. Revers transcription was performed by using 20 μg of RNA with Superscript RT III cDNA synthesis kit (Invitrogen). We performed quantitative real-time PCR using SYBR Green reagent (Bio-Rad, Hercules, CA). The cycle threshold values were obtained after normalizing to the nonreactive carboxy X-rhodamine dye, which served as a reference control. Using the δ-δ-cycle threshold method, after normalizing to glyceraldehyde-3-phosphate dehydrogenase (GAPDH), we quantified various transcripts using appropriate primers and analyzed melting curves to confirm accurate readings. The primer sequences are as follows: Gapdh forward: ATTGTTGCCATCAACGACCC; Gapdh reverse: CCACGACATACTCAGCACC; sterol regulatory element-binding protein (Srebp)-1c forward: TTCCAGAGAGGAGGCCAGAG; Srebp1c reverse: GGAGCCATGGATTGCACATT; acetyl CoA carboxylase 1 (Acc1) forward: GCGGGAGGAGTTCCTAATTC; Acc1 reverse: GGTTGGCATTGTGGATTTTC; fatty acid synthase (FAS) forward: CATGACCTCGTGAGAACGTGT; FAS reverse: CGGGTGAGGACGTTTACAAAG; hormone-sensitive lipase (HSL) forward: TCTAGCATGGGGTCCAGAGC; HSL reverse: TTCTGCGGCCTGGGAATTCC.

In vitro hepatocyte glucose secretion

Three weeks after treatment, livers from Ngn3-Btc-treated diabetic mice, empty vector-treated diabetic mice and nondiabetic control mice were perfusion digested by collagenase IV (Sigma) through the portal vein. We obtained a single cell suspension after filtration through a 100-μm mesh. Hepatocytes (5 × 10⁶) were plated in each of a six-well plate. The experiment was performed in duplicate using three mice from each group. The cells were washed in glucose-free KRB (Krebs Ringer Bicarbonate) buffer (119 mm NaCl, 4.7 mm KCl, 25 mm NaHCO₃, 2.5 mm CaCl₂, 1.2 mm MgSO₄, 1.2 mm KH₂PO₄ and 0.25% BSA) three times and incubated for 4 h in glucose-free KRBB. The glucose secretion into the media was measured using a glucose oxidase method (Infinity; Thermo, Waltham, MA).

Immunofluorescence and electron microscopy

We fixed liver and pancreas in 10% formalin overnight or perfusion fixed them with formaldehyde/gluteraldehyde solution and then processed tissues into 5-mm paraffin embedded sections. These were deparaffinized and hydrated using standard protocols with graded ethanol solutions. Antigen retrieval was performed using proteinase K, and primary antibody incubation with antiinsulin antibody was done overnight at 4 C, and the secondary antibody incubation (Alexa Fluor 488: goat antiguinea pig) was done at room temperature for 1 h. We mounted the slides with 4′,6′-diamino-2-phenylindole (DAPI) in the mounting medium and performed immunofluorescence microscopy using an Axiovert (Zeiss) microscope with Axiovision imaging software 4.0. All the images within the same experiment were acquired at the same settings and processed identically. For electron microscopy, the perfusion fixed tissues were embedded in LR white and sectioned. Ultrathin sections of 80 nm were stained with uranyl acetate and counterstained with lead citrate before visualization with an H-7500 transmission electron microscope (Hitachi, Indianapolis, IN).

LCM and global gene expression profiling

Liver and pancreas were embedded in optimum cutting temperature compound and frozen on dry ice and stored at −80 C. These were cryosectioned (7 μm) and nuclear stained (Histogene staining kit; Molecular Devices, Sunnyvale, CA) in an ribonuclease-free environment. We performed LCM using the Veritas microdissection system (Molecular Devices). We collected GFP-positive cells by LCM from the Ngn3-Btc-treated diabetic mip-GFP mice and also the liver and islets of nondiabetic mip-GFP mice. Three mice were used in each group. RNA was extracted, amplified, deoxyribonuclease I digested, and 20 μg reverse transcribed as described above. All the samples were labeled with Cy5 dye (Turbo labeling kit; Molecular Devices), whereas a pool from equal quantities of the three samples of normal islet RNA was labeled with Cy3 dye, which served as the reference sample. Equal quantities of Cy5 sample and Cy3 reference samples were mixed together and hybridized as per the manufacturer’s protocol to PancChip6.1 cDNA microarrays (β Cell Biology Consortium, University of Pennsylvania, Philadelphia, PA), scanned on an Agilent scanner (Palo Alto, CA), the images analyzed (Genepix pro 6.1; Molecular Devices), and the Cy5/to Cy3 ratios log2 transformed using median intensities, normalized by the Lowess normalization and sd regularization using MIDAS software (TIGR, The Institute for Genomic Research, Rockville, MD; Ref. 15). These data were transferred to MeV software (15) and a higher-order analysis using significance analysis of microarrays (SAM) (16) on genes that were expressed in 80% of the samples in the neo-islets and pancreatic islets. SAM was performed in MeV software with 500 random permutations, as a two-class unpaired testing using the K-nearest neighbor imputer with 10 neighbors in the imputing engine, S0 value set by the method of Tusher et al. (16), and δ value set at 0.4916.

Statistical methods

For all group comparisons, one-way ANOVA was performed with Tukey post hoc testing, and P ≤ 0.05 was considered significant.

Results

Ngn3-Btc treatment induces neo-islets and reverses hyperglycemia

Insulin-deficient diabetes was induced in C57BL/6 mice by administration of STZ, which manifested as stable hyperglycemia in these animals. Two weeks later, we treated the diabetic animals with a combination of HDAd-Ngn3 + HDAd-Btc (Ngn3-Btc) or empty vector. As we demonstrated previously (12), this treatment led to correction of hyperglycemia in a week (Fig. 1A); restoration of plasma insulin (Fig. 1B); and glucose tolerance (supplemental Fig. S1A, published as supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org), and cessation of weight loss (Fig. 1C-D), which continued to occur in mice treated with the empty vector. Because severe insulin deficiency is associated with unchecked lipolysis with consequent fat wasting, we measured the change in total body fat content with this treatment. Ngn3-Btc mice started gaining weight in 3 wk (Fig. 1C) due to an increase in total body fat (Fig. 1, E and F) as well as lean body mass (supplemental Fig. S2); they also looked much healthier compared with empty vector-treated animals.

Figure 1.

Ngn3-Btc treatment reverses hyperglycemia and restores body composition. Fasting blood glucose (A), fasting plasma insulin (B), body weight (C), percent change in body weight (D), percent body fat (E), and percent change in body fat proportion (F) in STZ-diabetic mice with indicated treatment and nondiabetic control mice are shown. Ngn3-Btc treatment restores all these parameters to that of nondiabetic controls (n = 6–7 in all experiments), and all values are mean ± sem. *, P ≤ 0.05 between Ngn3-Btc-treated vs. STZ-diabetic mice unless otherwise indicated.

The Ngn3-Btc mice remained euglycemic for the remainder of the study. The therapeutic efficacy of Ngn3-Btc was reproducible in multiple groups of STZ-diabetic mice. We verified that this was secondary to insulin production in the liver and the appearance of periportal neo-islets after Ngn3-Btc treatment (supplemental Fig. S1B).

Ngn3-Btc treatment reverses ketonemia

Insulin-deficient diabetic patients not only develop hyperglycemia, but they also show an increase in ketone bodies in the blood, which can lead to ketoacidosis, metabolic decompensation, and ultimately death. To address whether the insulin secreted by the neo-islets induced by the Ngn3-Btc treatment was effective in reversing this ketonemia, we measured the plasma level of β-hydroxybutyrate, a major ketone body in blood. As shown in Fig. 2A, the elevated plasma β-hydroxybutyrate level in empty vector-treated STZ diabetic mice was completely normalized in Ngn3-Btc-treated mice.

Figure 2.

Ngn3-Btc reverses ketonemia and normalizes glucose metabolism. A, Fasting plasma β-hydroxybutyrate in STZ-diabetic mice with indicated treatment and nondiabetic control mice (n = 3–4). B, In vitro glucose secretion from hepatocyte isolated from STZ-diabetic mice with indicated treatment and nondiabetic control mice (n = 3–4). C, Representative sections of mouse liver stained for glycogen with periodic acid-Schiff staining. Scale bar, 50 μm. D, Liver glycogen content in nondiabetic controls and STZ-diabetic mice with indicated treatment (n = 6–7) in all experiments, and all values are mean ± sem. *, P ≤ 0.05.

Ngn3-Btc treatment normalizes hepatic glucose secretion

Hepatic gluconeogenesis is the principal cause of fasting hyperglycemia, a cardinal feature of diabetes because insulin is the major physiological regulator of hepatic gluconeogenesis, and a loss of insulin action on the liver leads to unrestrained glucose production from the liver. We determined whether the reversal of hyperglycemia by Ngn3-Btc was in part due to the correction of the underlying insulinopenia-induced hepatic gluconeogenesis by directly measuring glucose secretion from hepatocytes isolated from the experimental animals. We reasoned that because hepatic glucose secretion is a net result of ongoing glycogenolysis and gluconeogenesis in the fasting state and because the Ngn3-Btc-treated mouse livers have more glycogen compared with diabetic mouse livers (see below and Fig. 2, C and D), any decrease in hepatic glucose secretion from Ngn3-Btc mouse hepatocytes would likely reflect a reversal of the unrestrained gluconeogenesis of the diabetic state.

We isolated hepatocytes from euglycemic nondiabetic control mice and euglycemic Ngn3-Btc-treated diabetic mice as well as hyperglycemic empty vector-treated diabetic mice 3 wk after treatment. We plated 5 × 10⁶ hepatocytes on 12-well plates and, after thorough washing with glucose-free KRBB buffer, measured the glucose secreted into the medium during a 4-h incubation. The amount of glucose secreted from hepatocytes of empty vector-treated diabetic mice was about 300% that of nondiabetic mice. In contrast, the glucose secretion from hepatocytes of Ngn3-Btc-treated mice was down-regulated to the same level as that of nondiabetic control (Fig. 2B). Therefore, Ngn3-Btc gene therapy completely corrected the elevated hepatic glucose secretion of the diabetic state.

Ngn3-Btc treatment restores hepatic glycogen

We next examined the effect of Ngn3-Btc treatment on hepatic glycogen content, which is the sum effect of two opposing insulin-regulated pathways: glycogenesis and glycogenolysis. In insulin-deficient diabetes, hepatic glycogen is depleted due to increased glycogenolysis and down-regulated glycogenesis. We stained liver sections from the three treatment groups for glycogen and found that it was abundant in liver sections from nondiabetic liver (Fig. 2C, left panel); in contrast, empty vector-treated diabetic liver sections were totally depleted of glycogen (Fig. 2C, middle panel). Ngn3-Btc treatment restored liver glycogen to a level similar to that of nondiabetic mice (Fig. 2C, right panel). A quantitative analysis of hepatic glycogen content confirmed that the livers of STZ-diabetic mice have significantly decreased glycogen content compared with nondiabetic mouse livers. With Ngn3-Btc therapy the hepatic glycogen content in the diabetic mice is restored to nondiabetic levels (Fig. 2D). Thus, Ngn3-Btc gene therapy replenished normal glycogen stores in the liver of diabetic mice.

Ngn3-Btc treatment normalizes disruptions of lipid metabolism in insulin-deficient diabetes

In addition to hyperglycemia, disruption of lipid metabolism contributes to the morbidity and mortality associated with patients with diabetes. We therefore determined whether, apart from normalizing glucose metabolism, Ngn3-Btc treatment also improved the dysregulated lipid metabolism in diabetic mice. In insulin-deficient diabetes, the lack of insulin stimulates lipolysis and down-regulates lipogenesis. We first measured plasma FFAs in these animals (Fig. 3A). Before induction of diabetes, all exhibited a similar plasma FFA level. Plasma FFA concentration was significantly elevated with onset of diabetes and in STZ-diabetic 6 wk after empty vector treatment. In contrast, Ngn3-Btc treatment restored the plasma FFAs to a level comparable with that of nondiabetic control mice. Furthermore, as shown in Fig. 3B, STZ-induced diabetes was associated with significant hypertriglyceridemia, which was also reversed by the Ngn3-Btc treatment to a level similar to that of nondiabetic mice. In addition, the severe catabolic state induced by the lack of insulin in the empty vector-treated diabetic mice was reflected by the almost complete absence of neutral lipid staining of liver sections from these animals, a phenotype that was also reversed upon Ngn3-Btc treatment, such that similar degrees of lipid staining were observed in liver sections of Ngn3-Btc-treated as in nondiabetic mice (Fig. 3C). A quantitative analysis of hepatic triglyceride content confirmed that the insulin-deficient diabetes-associated low hepatic triglyceride content is restored to that of nondiabetic levels with Ngn3-Btc therapy (Fig. 3D).

Figure 3.

Ngn3-Btc treatment normalizes hepatic lipid metabolism. Fasting plasma nonesterified fatty acid (NEFA) (A) and triglycerides (B) in STZ-diabetic mice with indicated treatment and nondiabetic control mice are shown. C, Representative sections of mouse liver stained for neutral lipid with Oil Red O. Scale bar, 20 μm. D, Liver triglyceride content in nondiabetic controls and STZ-diabetic mice with indicated treatment (n = 6–7 in all experiments), and all values are mean ± sem. *, P ≤ 0.05.

Ngn3-Btc treatment normalizes hepatic lipogenesis and lipolysis-related gene expression

To determine whether Ngn3-Btc treatment normalized lipogenesis and lipolysis at the gene transcript level, we measured the hepatic expression of the mRNA for key genes involved in lipogenesis and lipolysis. Srebp1c is a transcription factor and key regulator of lipogenesis that is regulated by the nutritional and hormonal state. The liver of diabetic mice treated with empty vector displayed a significantly down-regulated Srebp1c transcript level compared with that of nondiabetic control mice. This reduction in Srebp1c mRNA was completely reversed by Ngn3-Btc treatment. In keeping with this observation, the transcript level of two key enzymes involved in hepatic lipogenesis, ACC1 and FAS, was also restored to normal with Ngn3-Btc treatment, whereas empty vector treatment had no effect (Fig. 4, A–C). The expression of glycerol kinase, catalyzing an early step in triglyceride synthesis, was unchanged. In contrast to genes involved in lipogenesis, genes regulating lipolysis, such as HSL) and adipocyte triglyceride lipase are up-regulated in insulin-deficient STZ-diabetes and treatment with Ngn3-Btc restores the expression of these genes to that of nondiabetic controls (Fig. 4, C and D).

Figure 4.

Ngn3-Btc treatment normalizes hepatic lipogenesis and lipolysis-related gene transcripts. Quantitative RT-PCR for key lipogenesis and lipolysis-related gene transcripts in the liver from STZ-diabetic mice 6 wk after the indicated treatment and nondiabetic control mice are shown. Values are expressed after normalization relative to GAPDH (n = 6–7 in all experiments), and all values are mean ± sem. *, P ≤ 0.05. ATGL, Adipocyte triglyceride lipase.

Ngn3-Btc-induced neo-islets resemble normal pancreatic β-cells in their transcriptome

Ngn3-Btc treatment reverses hyperglycemia and other metabolic derangements in insulin-deficient diabetes for a lasting duration (over a year in multiple groups; experiments terminated before any of them developed recurrent hyperglycemia). This complete correction of metabolic derangements of insulin-deficient diabetes by Ngn3-Btc treatment requires that the induced neo-islets have all the components necessary for physiological regulation of insulin secretion. We reasoned that the neo-islets may have a transcription profile that is very similar to that of the native β-cells. To determine whether this is true, we performed a transcriptome analysis on RNA isolated by laser capture-microdissection from the neo-islets and from normal pancreatic islets. We first examined the microarray data for the four major islet hormones and found that the neo-islets induced in the liver by Ngn3-Btc treatment displayed similar levels of the transcripts for insulin, pancreatic polypeptide, and somatostatin but a reduced level for glucagon (Fig. 5A). Quantitative PCR analysis of LCM isolated RNA of the islet-related genes, such as Ins1 and glucagon genes, confirmed the microarray data (Fig. 5B). In addition, genes such as Pdx1, Glut2, and Ffar1 (GPR40), important for appropriate GSIS (5,17,18), were found to be expressed at similar levels in the neo-islets and pancreatic islets (Fig. 5B). To define the degree of similarity or difference at a global level between the Ngn3-Btc-induced neo-islets and native pancreatic islets, we analyzed the microarray data using SAM on the gene transcripts and found the transcriptome of the neo-islets to be remarkably similar to that of native pancreatic islets (Fig. 5C). The scatter plot in Fig. 5C represents the expected (x-axis) and observed (y-axis) relative values for all the genes that are expressed in at least 80% of the samples (5289 genes). The interrupted lines represent the ±δ-value from the solid line, which represents the observed = expected line. The genes that fall within the interrupted lines are considered not significant. In this comparison of neo-islets and pancreatic islets, virtually none of the genes display any significant difference between the two groups. Even with a small δ-value (δ = 0.4916), there is a remarkable lack of difference between the two groups, i.e. the transcriptome of the periportal neo-islets is very similar to that of native pancreatic islets.

Figure 5.

Ngn3-Btc-induced neo-islets express individual islet hormones. A, Microarray-derived expression of pancreatic hormone transcripts from Ngn3-Btc induced neo-islets compared with that of native pancreatic islets. B, Quantitative RT-PCR for islet-related genes from LCM-obtained RNA from neo-islets from the liver of Ngn3-Btc-treated diabetic mice and from nondiabetic control pancreatic islets. Values are expressed after normalization relative to GAPDH (n = 3), and all values are mean ± sem. C, A SAM plot showing the close similarity of the expressed transcript profiles of Ngn3-Btc-induced neo-islets and normal pancreatic islets (see text for details). Ins1, Insulin1; Ins2, insulin2; gcg, glucagon; ppy, pancreatic polypeptide; sst, somatostatin; Glut2, glucose transporter 2; Pdx1, pancreatic and duodenal homeobox 1 (Ipf1); Ffar1, FFA receptor 1 (GPR40).

Ngn3-Btc-induced neo-islets contain dense-core granules typical of normal pancreatic β-cells

We examined by electron microscopy the ultrastructure of the periportal cells induced by Ngn3-Btc and compared them with that of normal pancreatic β-cells. Normal mouse pancreatic β-cells are characterized by the presence of electron-dense core granules that are surrounded by a halo inside the secretory vesicle (Fig. 6, right panel). The Ngn3-Btc induced β-cell-like cells in the periportal region of the liver demonstrated a similar morphology of electron dense-secretory granules surrounded by a halo (Fig. 6, left panel).

Figure 6.

Ngn3-Btc-induced neo-islets have an ultrastructure similar to pancreatic β-cells. Electron micrograph of an induced neo-islet in the periportal region from a Ngn3-Btc-treated mouse liver (left panel). Arrows, Electron-dense granules with a halo typical of mature insulin secretory granules seen in a normal pancreatic islet (right panel). PV, Portal vein.

Discussion

Diabetes is a metabolic disorder that results from derangements in glucose and lipid homeostasis (19,20). Insulin deficiency is associated with a raging catabolic state with consequent peripheral lipolysis, impaired hepatic lipogenesis, ketogenesis, and disrupted metabolite cycling in the liver. Achieving transient euglycemia is not necessarily associated with complete metabolic correction. Because of the difficulty of inducing GSIS and euglycemia in experimental diabetes gene therapy, all publications on the topic to date have focused entirely on whether the treatment leads to a return of euglycemia in response to the treatment strategy (reviewed in Refs. 10, 11). The Ngn3-Btc gene transfer has enabled us to restore euglycemia and return of GSIS. We reasoned that an effective antidiabetic treatment strategy must also have the capacity to normalize whole-body metabolic homeostasis to achieve the maximal benefit and forestall the acute and chronic complications that may ensue.

We addressed this in our analysis, which showed that Ngn3-Btc-induced euglycemia in STZ-diabetic mice is associated with reversal of multiple abnormal lipid parameters, including plasma ketones, FFAs, and triglycerides. In addition, the restoration of lean mass with Ngn3-Btc therapy also indicates a cessation of the severe catabolic processes associated with insulin-deficient diabetes and a restoration of a positive nitrogen balance with resultant protein synthesis and increased muscle mass (supplemental Fig. S2). To date, the Ngn3-Btc-induced islet neogenesis approach appears to be the only strategy for total normalization of multiple parameters of glucose and lipid metabolism in STZ mice by gene therapy (reviewed in Refs. 10, 11, and 21). The vast majority of previous reports on diabetes gene therapy used hepatic insulin gene transfer as the form of treatment. In instances in which such parameters were measured, this latter strategy was associated with significant increases in serum FFAs and triglycerides (2,3,6) compared with nondiabetic controls. These complications appear to be related to the relatively high concentration of insulin in the transduced hepatocytes leading to an abnormal activation of the lipogenic process that results in elevated serum FFAs and triglycerides, which are not observed in the Ngn3-Btc-induced islet neogenesis approach used in this study.

To achieve this level of metabolic correction, the Ngn3-Btc-induced neo-islets need to acquire the GSIS machinery of pancreatic β-cells, specifically its stimulus-secretion coupling of insulin secretion (5,22). Primed insulin containing vesicles in the β-cell are a necessary component for this rapid phase of stimulus-secretion coupling and usually represent vesicles that are ready for immediate exocytosis. These vesicles are seen as electron dense core secretory granules on electron microscopy. The presence of these granules indicates a level of maturity in these β-cells that can only occur resulting from a supporting network of many genes involved in transcription, translation, posttranslational processing, and packaging with other chaperone proteins. We performed a global transcriptome analysis and showed the remarkable degree of similarity of the neo-islet transcriptome to that of the native β-cell. Other therapies that lead to constitutive production of insulin from cells that lack the repertoire of β-cell GSIS machinery may still be sufficient to reverse hyperglycemia under basal conditions but have not been shown to display similar dense core granules. This lack of mature granules often results in impaired GSIS and an incomplete correction of the metabolic derangements of diabetes. Somatic gene therapy using the insulin transgene leads to the production of insulin via the constitutive pathway and is not associated with insulin granules packaged in secretory vesicles (22). In this study we performed transmission electron microscopy to examine the ultrastructure of the Ngn3-Btc-induced neo-islets and found β-cell-like cells in the periportal region of the liver of treated animals. The presence of the characteristic vesicles containing secretory granules adorned by a halo in these cells suggests that the cells may have attained a degree of maturity that allow them to display GSIS as well as correct the multiple metabolic derangements that occur in insulin-deficient diabetes.

In conclusion, the treatment of diabetes with the Ng3-Btc regimen leads to the induction of hepatic periportal neo-islets that resemble β-cells in both ultrastructure and function, resulting in a complete reversal of all the major metabolic disruptions observed in untreated insulin-deficient STZ-diabetic mice.