Insulin Null β-cells Have a Prohormone Processing Defect That Is Not Reversed by AAV Rescue of Proinsulin Expression

Abstract

Up to 6% of diabetes has a monogenic cause including mutations in the insulin gene, and patients are candidates for a gene therapy. Using a mouse model of permanent neonatal diabetes, we assessed the efficacy of an adeno-associated virus (AAV)-mediated gene therapy. We used AAVs with a rat insulin 1 promoter (Ins1) regulating a human insulin gene (INS; AAV Ins1-INS) or native mouse insulin 1 (Ins1; AAV Ins-Ins1) to deliver an insulin gene to β-cells of constitutive insulin null mice (Ins1^−/−Ins2^−/−) and adult inducible insulin-deficient mice [Ins1^−/−Ins2^f/f PdxCreER and Ins1^−/−Ins2^f/f mice administered AAV Ins1-Cre)]. Although AAV Ins1-INS could successfully infect and confer insulin expression to β-cells, insulin null β-cells had a prohormone processing defect. Secretion of abundant proinsulin transiently reversed diabetes. We reattempted therapy with AAV Ins1-Ins1, but Ins1^−/−Ins2^−/− β-cells still had a processing defect of both replaced Ins1 and pro-islet amyloid polypeptide (proIAPP). In adult inducible models, β-cells that lost insulin expression developed a processing defect that resulted in impaired proIAPP processing and elevated circulating proIAPP, and cells infected with AAV Ins1-Ins1 to rescue insulin expression secreted proinsulin. We assessed the subcellular localization of prohormone convertase 1/3 (PC1/3) and detected defective sorting of PC1/3 to glycogen-containing vacuoles and retention in the endoplasmic reticulum as a potential mechanism underlying defective processing. We provide evidence that persistent production of endogenous proinsulin within β-cells is necessary for β-cells to be able to properly store and process proinsulin.

Diabetes affects over 537 million people worldwide, and although the majority of patients have either type 1 or type 2 diabetes (1), 1% to 6% of all cases of diabetes have a monogenic cause (2). There have been at least 22 genes identified as causative for monogenic diabetes (3, 4) including transcription factors HNF1α (MODY3), NEUROD1 (MODY6), and PAX4 (MODY9), genes regulating β-cell function such as GCK (MODY2) and KCNJ11 (permanent and transient neonatal diabetes) and even insulin (INS) itself (MODY10 and permanent neonatal diabetes) (5). Monogenic causes of diabetes provide valuable insight into the pathogenesis of more common type 2 diabetes because of the known associations between monogenic diabetes gene variants and type 2 diabetes (5).

Mutations in the insulin gene (INS) often cause neonatal diabetes via multiple pathways. Insulin is made from a precursor called proinsulin. Mouse proinsulin 1 and proinsulin 2 require cleavage at specific internal sites by prohormone convertase 1/3 (PC1/3) and 2 (PC2) (6) whereas human β-cells rely predominantly on PC1/3 to process proinsulin (7). Correct processing, disulphide bond formation, and folding are essential for production of mature insulin with full bioactivity. More than 20 mutations have been detected and found to be causative for permanent neonatal diabetes (8) and occasionally mutant INS gene-induced diabetes of youth (9). Based on work in the Akita mouse [C96Y mouse insulin 2 (Ins2)], the best available evidence suggests that insulin mutations cause diabetes via mutant proinsulin acting in a dominant negative fashion wherein mutated proinsulin misfolds and forms aggregates that incorporate the bystander normal insulin. Most causative mutations involve cysteine residues that alter disulfide bonds, protein folding, and proinsulin maturation (10), leading to a buildup of misfolded aggregates in the endoplasmic reticulum (ER) that limits β-cell expansion in neonates (11), and β-cell death from ER stress (12).

A mouse model of permanent neonatal diabetes with deletion of both nonallelic insulin genes (Ins1^−/−Ins2^−/−) has been generated (13). Of note, though complete loss of the insulin gene does not fully mimic the more common etiology of mutant INS gene-induced diabetes of youth, loss of normal insulin production does occur in rarer forms (14). Additionally, Ins1^−/−Ins2^−/− mice have dedifferentiated β-cells that lack mature β-cell factors including V-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MAFA) and homeobox protein Nkx6.1 (NKX6.1) (15). After birth, Ins1^−/−Ins2^−/− pups become hyperglycemic, and replacement of insulin by injection was not sufficient for the maturation of insulin-deficient β-cells (15). In contrast, long-term replacement of insulin by islet transplantation better supported endogenous β-cells based on them gaining immunoreactivity for mature β-cell factors including MAFA (15). Assessing the viability of a gene therapy approach to treat this model of permanent neonatal diabetes could provide insight into potential treatments for patients with diabetes caused by mutations in the INS gene.

The eighth serotype of the adeno-associated virus (AAV8) is a candidate gene therapy vector for treating monogenic diabetes because AAV8 has high affinity for pancreatic islets (16) and AAVs are a clinically relevant therapy. In the current study, we aimed to determine whether AAV8 carrying an insulin open reading frame regulated by an insulin promoter (Ins1) could restore insulin production to Ins1^−/−Ins2^−/− β-cells and resolve diabetes. Here we show that although AAVs delivered either human (AAV Ins1-INS) or mouse (AAV Ins1-Ins1) insulin genes to β-cells, defects in proinsulin and proIAPP processing endured in Ins1^−/−Ins2^−/− β-cells, which prevented them from forming normal dense core secretory granules. Adult inducible insulin-deficient mouse models had defective processing of proIAPP due to defective sorting of PC1/3. Our findings support the capability of the AAV8 to deliver the insulin gene to pancreatic β-cells, but its efficacy at reversing diabetes in mice is limited by an persistent prohormone processing defect in insulin-deficient β-cells.

Materials and Methods

Animal Models and Insulin Therapy

All experiments were approved by the UBC Animal Care Committee and carried out in accordance with the Canadian Council on Animal Care Guidelines. We present findings from a mix of male and female animals in all experiments. Animals were given ad libitum access to a standard chow diet (2918, Harlan Laboratories, Madison, WI, USA) and housed with a 12-hour light/dark cycle. Generation of Ins1^−/−Ins2^−/− mice and insulin therapy were performed as previously described (15). Animals received twice-daily subcutaneous injections of approximately 0.1U insulin glargine (Lantus®; diluted to 5 U/mL in F-10 media, Sigma-Aldrich, St. Louis, MO, USA) until 15 days of age when they were either treated with islet transplantation into the anterior chamber of the eye (17) or with insulin AAV and insulin therapy continued as needed. Blood glucose monitoring was done by random-fed measurements or after a 4-hour fast (9 am-1 pm) using a One Touch Ultra glucometer (Life Scan Inc., Burnaby, Canada). For the intraperitoneal glucose tolerance test (IPGTT) and fast/refeed, mice were fasted for 4 hours (9 am-1 pm) before glucose injection (1-2 g glucose/kg body weight IP) or return of food. For insulin tolerance tests, mice were administered Novolin® at a dose of 0.8 U/kg body weight IP. AAVs were delivered by IP injection. Further details are in the Supplemental Methods (18).

Adeno-associated Virus Production

All plasmids used in this study were generated from the double-stranded AAV mINS2p-EGFP plasmid, kindly provided by Dr. Paul Robbins (16). After replacement of the promoter and open reading frame with targets of interest (Ins1-GFP, Ins1-INS, Ins1-Cre, and Ins1-Ins1), plasmids were sent to either the Children’s Hospital of Philadelphia (Philadelphia, PA, USA; Ins1-Cre), SAB Technology Inc. (Philadelphia, PA, USA; Ins1-INS), or Vector BioLabs (Malvern, PA, USA) to manufacture high-titer double-stranded AAV. Further details are in the Supplemental Methods (18).

Islet Isolation and Transplantation

Ins1^−/−Ins2^+/− mice were euthanized, and pancreatic islets were isolated by collagenase digestion (1000 U/mL type XI collagenase; Sigma-Aldrich, St. Louis, MO, USA) (19) and transplanted into the anterior chamber of the eye (17).

Assays

We assessed circulating plasma levels of human C-peptide (Mercodia cat no. 10-1141-01, Uppsala, Sweden), mouse C-peptide (Alpco cat no. 80-CPTMS-E01, Salem, NH, USA), and proinsulin (human: Mercodia cat no. 10-1118-01, Uppsala, Sweden; mouse: Mercodia cat no. 10-1232-01) by commercially available enzyme-linked immunosorbent assay [cross-reactivities in Supplemental Table 1 (18)]. For detection of insulin autoantibodies, plasma was sent to the Barbara Davis Center (University of Colorado, Aurora, CO, USA) for a micro insulin autoantibody radiobinding assay. Human insulin was used as the antigen thus allowing detection of mouse immunoglobulin G specific for epitopes of human insulin.

Immunohistofluorescence and Immunohistochemistry

Paraformaldehyde-fixed pancreases were immunostained by standard protocol as previously described [primary antibodies, including RRIDs, are listed in Supplemental Table 2 (18)] (15). Further details are in the Supplemental Methods (18).

Western Blots

Plasma collected from Ins1^−/−Ins2^f/fPdxCreER mice was used for a tris-tricine (20) fluorescent Western blot using an IAPP antibody (1:500, Ab15125, AbCam) that was detected and analyzed on a LI-COR Odyssey 9120 Imaging system (LI-COR Biosciences, Lincoln, NE, USA). Further details are in the Supplemental Methods (18).

Transmission Electron Microscopy

Paraformaldehyde or glutaraldehyde samples were sent to the Canadian Center of Electron Microscopy at McMaster University (Hamilton, Canada) for sample preparation and imaging by standard methods. Further details are in the Supplemental Methods (18).

Statistical Analysis

Data were subject to the Shapiro-Wilk normality test, and when all groups passed the test for normality, we analyzed by parametric test. We used Student’s t-test to compare 2 groups (paired or unpaired), 1-way analysis of variance with Tukey test for multiple comparisons, and 2-way analysis of variance with Bonferroni’s post hoc testing. When any group failed the test for normality, we used the nonparametric Mann-Whitney U test or the Kruskal-Wallis test with Dunn’s multiple comparisons. Statistical analysis was performed using GraphPad Prism 8.00 (La Jolla, CA, USA) with significance set at P < 0.05.

Results

AAV Ins1-INS Can Produce Dose-dependent Infection of Mouse β-cells

We injected wild-type Ins1^+/+Ins2^+/+ mice with variable doses of AAV (in all studies, AAV was delivered by IP injection) or saline and collected pancreases 42 days later. We immunostained for IAPP as a marker of β-cells [some delta cells have IAPP (21, 22)] and assessed infection rate by immunostaining for human C-peptide. Using human pancreases as a positive control, we detected a dose-dependent increase in the proportion of infected β-cells (Fig. 1A). We performed an IPGTT on day 20 following AAV delivery (2 g glucose/kg body weight) and assayed for mouse C-peptide and human C-peptide by commercial enzyme-linked immunosorbent assay. Mouse C-peptide was stable (Fig. 1B), whereas human C-peptide immunoreactivity increased in a dose-dependent manner (Fig. 1C). Serial IPGTT were performed on days −1, 13, and 40 relative to AAV administration, and there was a transient improvement in glucose tolerance in the 2 higher-dosed groups (Fig. 1D). Given that peak AAV expression is known to occur approximately 2 weeks post-AAV injection (16), we hypothesized that the transient improvement was attributable to insulin expression and production off of the AAV-delivered plasmid.

Figure 1.

β-cells produce human proinsulin after delivery of adeno-associated virus (AAV) insulin 1 (Ins1)-INS in vivo. Pancreas collected 6 weeks after 8-week-old wild-type mice (n = 3 per group) were treated with AAV Ins1-INS immunostained for human C-peptide and islet amyloid polypeptide (IAPP) (A). Human pancreases were used as positive control (representative images of n = 2-3; scale bar is 100 μm). Quantification of the proportion of IAPP + cells immunoreactive for human C-peptide shown to the right. Plasma collected during an intraperitoneal glucose tolerance test (IPGTT; 2 g glucose/kg body weight) on day 20 relative to AAV Ins1-INS injection was assayed for mouse C-peptide (B) and human C-peptide (C). (D) IPGTTs on days −1, 13, and 40 relative to AAV Ins1-INS treatment. (E) Pancreases were immunostained for insulin (INS) and proinsulin specific antibodies that bind to the unprocessed B/C junction or unprocessed C/A junction of proinsulin. Representative images of n = 3, insets are enlarged 4×, and the scale bar is 100 µm. Data presented on box-and-whisker plots (A), mean in bold with faint lines for individual mice (B and C), or mean ± SE of the mean (D). All groups compared by 1-way (A) or repeated measures 2-way analysis of variance (B-D). *P < 0.05, **P < 0.01, ***P < 0.001. Abbreviation: VGP, viral genome particles.

To verify that AVV Ins1-INS transduced wild-type β-cells had normal proinsulin processing, we immunostained with antibodies specific for an unprocessed proinsulin junction using an antibody that binds to an intact C-A junction (C/A) and an antibody that binds to the intact B-C junction (B/C) (6). Use of these B/C and C/A antibodies to assess proinsulin processing in mice has been well characterized with clear ability to differentiate processing or failed processing at both cleavage sites based on careful studies with PC1/3 and PC2 null mice (6). In saline- and AAV-treated β-cells, there was an expected perinuclear pattern of immunoreactivity (Fig. 1E). The perinuclear pattern reflects normal processing with some proinsulin in the perinuclear ER compartment (proinsulin is translated but not yet processed within the ER), but no proinsulin immunoreactivity in the cell periphery where the mature secretory granules are found.

AAV Ins1-INS Transiently Reverses Diabetes in Ins1^−/−Ins2^−/− Mice

We generated a cohort of Ins1^−/−Ins2^−/− pups and determined whether treatment with AAV Ins1-INS could reverse diabetes in this model of neonatal diabetes. Using Ins1^−/−Ins2^+/− littermates and wild-type Ins1^+/+Ins2^+/+ mice as controls, body weight (Fig. 2A) and random blood glucose (Fig. 2B) was monitored from birth for 7 weeks. Ins1^−/−Ins2^−/− pups were treated with insulin glargine (Lantus®) 2× or 3× daily until 15 days of age when groups were injected with either 5 × 10¹⁰ viral genome particles (VGP) AAV Ins1-INS/g body weight or saline. Ins1^−/−Ins2^−/− mice became insulin independent 1 week after AAV treatment [Supplemental Figure 1A (18)] but were glucose intolerant at 28 days of age [Supplemental Figure 1B (18)]. At 28 days of age, Ins1^−/−Ins2^−/− mice had no detectable circulating mouse C-peptide (Fig. 2C) but had detectable circulating human C-peptide (Fig. 2D). AAV Ins-INS–treated controls also had detectable human C-peptide, but at levels lower than Ins1^−/−Ins2^−/− mice. Given variable glucose tolerance [Supplemental Figure 1B (18)] and that the area under the curve of glucose during the IPGTT at 28 days of age was inversely correlated with levels of fasting plasma human C-peptide (b = −353.8, SE = 142.3, P < 0.05, 95th CI −56.03, −651.6) [Supplemental Figure 1C (18)], we hypothesized that low rates of β-cell infection contributed to relapse to diabetes. Thus, the 21 Ins1^−/−Ins2^−/− mice were divided into 3 groups: (1) worst glucose tolerance (AAV/−; n = 11) or better glucose tolerance (2) AAV/AAV (n = 5), and (3) AAV/saline (n = 5). The AAV/− group was not retreated, the AAV/AAV group was reinjected with AAV Ins1-INS, and the AAV/saline group was injected with saline (to serve as a direct control group for the AAV retreated group). Despite retreatment at 5 weeks of age (AAV/AAV group) and severe relapse to diabetes at ~4 to 5 weeks of age, all groups of Ins1^−/−Ins2^−/− animals had comparable levels of human C-peptide at 4, 6, and 8 weeks of age [Supplemental Figure 1D (18)].

Figure 2.

Transient remission of diabetes in Ins1^−/−Ins2^−/− mice after treatment with adeno-associated virus (AAV) insulin 1 (Ins1)-INS. Random fed body weight (A) and blood glucose (B) of Ins1^−/−Ins2^−/−, Ins1^−/−Ins2^+/− littermates, and Ins1^+/+Ins2^+/+ control mice monitored from birth. Ins1^−/−Ins2^−/− animals were grouped as nonresponders with the worst glucose tolerance (designated as AAV/−) and responders with better glucose tolerance during intraperitoneal glucose tolerance test at 28 days of age (responders were retreated at 5 weeks of age with AAV or saline and are designated as AAV/AAV or AAV/saline, respectively). Four-hour fasted plasma collected at 28 days of age was assayed for mouse C-peptide (C) and human C-peptide (D) (a, b, and c represent statistical differences between groups, P < 0.05). Plasma collected after a 4-hour fast was used in an assay specific for proinsulin and proinsulin intermediates (E) (18/21 Ins1^−/−Ins2^−/− mice had detectable proinsulin). Pancreas was immunostained for islet amyloid polypeptide (IAPP) and human C-peptide (F) or human C-peptide, mouse C-peptide, and either an unprocessed C-A junction proinsulin and/or unprocessed B-C junction proinsulin (G) (6). Quantification of the proportion of human or mouse C-peptide immunoreactive area that colocalizes with immunoreactivity for a cut site–specific antibody, and the average intensity of the cut site–specific antibody immunoreactivity in the C-peptide immunoreactive area shown on the right. Pancreas was immunostained for insulin and either an unprocessed C-terminus or N-terminus of IAPP as well as IAPP and a mature insulin specific antibody (H). Pancreas was immunostained for critical prohormone-processing enzymes prohormone convertase (PC) 1/3, PC2, the PC2 cofactor neuroendocrine protein 7B2, and carboxypeptidase E (I). Representative images of n = 4-5, scale bars are 100 μm, and insets are enlarged 4×. Data presented as mean ± SE of the mean (A and B) or box-and-whisker plots with individual animals shown as data points (C-E). All groups were compared by repeated measures 2-way (A and B) or 1-way analysis of variance (C-E). *P < 0.05, **P < 0.01.

To reconcile the discordant observations of abundant human C-peptide yet severe hyperglycemia, we assayed plasma for neutralizing insulin autoantibodies, but there was no elevation of insulin autoantibodies in AAV Ins1-INS–treated animals [Supplemental Figure 1F (18)]. Given the published cross-reactivity of the human C-peptide assay with proinsulin [Supplemental Table 1 (18)], we used an assay specific for proinsulin and the proinsulin intermediates (Mercodia, 10-1118-01) and observed that there was circulating proinsulin in 18 of 21 AAV-treated Ins1^−/−Ins2^−/− mice (Fig. 2E). Proinsulin has some bioactivity [Supplemental Table 3 (18)] that may lower blood glucose in highly insulin-sensitive pups but may not be adequate in larger mice reaching adulthood. We performed immunohistofluorescence experiments and confirmed the presence of transduced β-cells with human C-peptide immunoreactivity (Fig. 2F), with rates of infection of approximately 5% in the AAV/− group and approximately 10% in other groups [Supplemental Figure 1E (18)]. Although the proportion of β-cells with human C-peptide immunoreactivity was only 5% to 10% in the AAV-treated pups, the actual infection rate at the time of administration of the AAV at 15 days of age may have been closer to the 68% observed in control mice (Fig. 1A), but β-cell replication near the time of weaning (23, 24), β-cell neogenesis (25-27), and/or degradation of viral genomes diluted the proportion of β-cells with C-peptide immunoreactivity.

To assess proinsulin processing in successfully transduced cells, we immunostained with antibodies specific for the B/C and C/A junctions of proinsulin (6). In Ins1^−/−Ins2^−/− pancreases, we observed pancytoplasmic immunoreactivity using the B/C junction antibody that was more intense than in controls and colocalized with the majority of the human C-peptide immunoreactive area, suggesting impaired B/C site processing of AAV Ins1-INS–infected Ins1^−/−Ins2^−/− β-cells (Fig. 2G). Furthermore, we assessed the processing of proIAPP by immunostaining for the C- and N-terminal fragments of unprocessed IAPP. In mice, proIAPP is processed by PC1/3 and then PC2 sequentially at the C-terminus and then the N-terminus, respectively (28, 29). proIAPP immunoreactivity was increased in both infected (with human C-peptide immunoreactivity) and uninfected Ins1^−/−Ins2^−/− β-cells, as well an absence of immunoreactivity for mature insulin (Fig. 2H). Surprisingly, Ins1^−/−Ins2^−/− β-cells had normal immunoreactivity for the proinsulin processing enzymes PC1/3, PC2, 7B2, and carboxypeptidase E (Fig. 2I; Supplemental Figure 1G and 1H (18)]. Based on previous observations that most Ins1^−/−Ins2^−/− β-cells stop expressing beta galactosidase after birth (the Lacz gene is knocked in to the Ins2 locus) (13), we immunostained pancreases for human C-peptide and beta galactosidase but found little to no colocalization [Supplemental Figure 1 (18)], suggesting that the observed low rate of human C-peptide immunoreactive cells was not due to silencing of the insulin promoter in AAV Ins1-INS–transduced β-cells. We also immunostained for markers of β-cell maturity, GLUT2 and NKX6.1, and found normal nuclear NKX6.1 but a lack of membranous GLUT2 in hyperglycemic AAV-treated Ins1^−/−Ins2^−/− mice [Supplemental Figure 1I (18)], suggesting cells were functionally mature but GLUT2 deficient due to chronic hyperglycemia (30, 31).

Ins1^−/−Ins2^−/− β-cells Are Unable to Process Mouse Insulin 1 but High-dose AAV Can Reverse Diabetes

Given that Ins1^−/−Ins2^−/− β-cells were unable to process human insulin delivered via AAV Ins1-INS, we aimed to determine whether they could process native mouse proinsulin. Given low rates of infectivity after treating mice at 15 days of age (approximately 5-10%), we treated Ins1^−/−Ins2^−/− mice with islet transplants to keep them alive until postweaning before treating with AAV to have a larger target β-cell mass. We tracked blood glucose (Fig. 3A) of Ins1^−/−Ins2^−/− mice, Ins1^+/+Ins2^+/+ wild-type controls, and Ins1^−/−Ins2^+/− littermate controls. Four Ins1^−/−Ins2^−/− mice were given isogenic islet transplants into the anterior chamber of the eye at 15 days of age, but 2 mice died during or shortly after surgery, so we present data from the remaining n = 2 (group denoted Ins1^−/−Ins2^−/− islets + AAV). The Ins1^−/−Ins2^−/− islets + AAV mice were treated with high-dose (5 × 10¹¹ VGP/g body weight) AAV Ins1-Ins1 at 30 days of age, and the islet grafts were removed by enucleation at 49 days of age. Plasma collected at 8 and 10 weeks of age was assayed for C-peptide, and Ins1^−/−Ins2^−/− islets + AAV mice had normal C-peptide immunoreactivity (Fig. 3B), but this was likely largely attributable to cross-reactivity to extreme hyperproinsulinemia (Fig. 3C). We collected pancreases at 11 weeks of age and observed high rates of transduction (63.7 ± 7.7%) [Supplemental Figure 2A (18)].

Figure 3.

Ins1^−/−Ins2^−/− β-cells have a prohormone processing defect. Ins1^−/−Ins2^−/− pups were treated with insulin injections (0-15 days of age) and then transplanted with islets into the anterior chamber of the eye at 15 days of age (triangle), treated with 5 × 10¹² viral genome particles adeno-associated virus (AAV) insulin 1 (Ins1)-Ins1 at 28 days of age (arrow) and enucleated at 49 days of age; their blood glucose was monitored from birth to 70 days of age (A). Plasma was collected after enucleation at 8 and 10 weeks of age and assayed for mouse C-peptide (B) and proinsulin (C). Mice were euthanized at 11 weeks of age, and pancreases were immunostained using a pan-insulin antibody (green) and prohormone-specific antibodies targeting the unprocessed B/C junction of proinsulin, C/A junction of proinsulin, N-terminus of pro-islet amyloid polypeptide (proIAPP), or the C-terminus of proIAPP (D). Representative image of n = 2-3. Scale bar is 100 µm. Transmission electron microscopy of β-cells in wild-type Ins1^+/+Ins2^+/+ and AAV Ins1-Ins1–infected endogenous β-cells (E). Representative images of n = 2, the scale bar is 3 μm, and insets are enlarged 4×. Immunostaining of pancreases from Ins1^+/+Ins2^+/+ controls and AAV-treated Ins1^−/−Ins2^−/− mice for evidence of AAV infection (C-peptide 1) and for markers of β-cell maturity including MAFA, NKX6.1, NKX2.2, PDX1, and GLUT-2 (F). Representative images of n = 2-3. Scale bar is 100 μm. Groups were compared by repeated measures 2-way (A) or 1-way analysis of variance with Tukey’s post hoc test (B and C). *P < 0.05, **P < 0.01.

We assessed proIAPP and proinsulin processing in β-cells of wild-type and islets + AAV-treated Ins1^−/−Ins2^−/− mice. We observed pancytoplasmic immunoreactivity with the N-terminal proIAPP antibody in Ins1^−/−Ins2^−/− β-cells, suggesting poor N-terminal proIAPP processing (Fig. 3D) and bright cytoplasmic immunoreactivity for the unprocessed C/A junction of proinsulin (Fig. 3D). There was immunoreactivity for PC1/3, PC2, 7B2, and carboxypeptidase E in Ins1^−/−Ins2^−/− β-cells [Supplemental Figure 2B (18)], but transmission electron microscopy on pancreas sections revealed that, instead of dense core secretory granules seen in wild-type mice, there were only immature low-density granules lacking halos in infected β-cells of Ins1^−/−Ins2^−/− islets + AAV mice (Fig. 3E). Illustrative counting of a representative β-cell in Ins1^−/−Ins2^−/− islets + AAV mice (n = 2) revealed that 94% to 97% of granules had an immature phenotype whereas in wild-type mice (n = 2) only 8% to 11% of granules had an immature phenotype. Examination of neonatal pancreases from Ins1^−/−Ins2^−/− identified cells devoid of mature granules that instead contained only very few immature-appearing granules that likely contained IAPP [Supplemental Figure 4 (18)]. Findings of immature granules, cytoplasmic immunoreactivity for prohormones (Fig. 3D), and elevated circulating proinsulin (Fig. 3C) suggest that proinsulin processing is markedly impaired in AAV Ins1-Ins1–infected Ins1^−/−Ins2^−/− β-cells. We immunostained pancreases from AAV Ins1-Ins1–treated mice and controls for markers of β-cell maturity and dedifferentiation. Ins1^−/−Ins2^−/− islets + AAV pancreases had normal nuclear immunoreactivity for PDX1, NKX2.2, NKX6.1, and MAFA, as well as GLUT2 localized to the plasma membrane (Fig. 3F). Immunostaining for progenitor markers that have been used as β-cell dedifferentiation markers (15, 32), including NANOG, L-MYC, and ALDH1A3, provided no evidence for β-cell dedifferentiation [Supplemental Figure 2C (18)].

Defective PC1/3 sorting underlies persistent prohormone processing defects in insulin-deficient mice transduced to express proinsulin.

Given the persistent severe prohormone processing defect of constitutive insulin-deficient β-cells in Ins1^−/−Ins2^−/− mice, we attempted to prevent diabetes onset in adult male Ins1^−/−Ins2^f/f mice (6-8 weeks of age) by treating them with 1.5 × 10¹² VGP AAV Ins1-Cre with or without 2.5 × 10¹² VGP AAV Ins1-Ins1, or phosphate-buffered saline (PBS). There were no differences in body weight over the duration of the study [Supplemental Figure 3A (18)], and both AAV-treated groups developed fasting hyperglycemia 4 weeks after infection (Fig. 4A). We assayed plasma for C-peptide (Fig. 4B) and observed no significant differences between groups whereas proinsulin (Fig. 4C) was transiently increased in the AAV Ins1-Cre + AAV Ins1-Ins1–treated group. AAV-treated groups had similar glucose intolerance during IPGTTs (2 g glucose/kg body weight) 22 days [Supplemental Figure 3B (18)] and 46 days [Supplemental Figure 3C (18)] post-AAV compared to controls. As a potential explanation for observations of glucose intolerance but nonsignificant changes to circulating C-peptide levels, we performed an insulin tolerance test on day 51 relative to AAV to assess insulin sensitivity [Supplemental Figure 3E (18)] but observed no significant differences in blood glucose.

Figure 4.

Adult inducible insulin-deficient β-cells develop a prohormone processing defect. Fasting blood glucose (A) of inducible insulin knockout Ins1^−/−Ins2^f/f mice that received adeno-associated virus (AAV) insulin 1 (Ins1)-Cre, AAV Ins1-Cre plus AAV Ins1-Ins1, or phosphate-buffered saline by intraperitoneal injection on day 0. Plasma was collected after a 4-hour fast and assayed for C-peptide (B) and proinsulin (C). Thirteen weeks post-AAV, pancreases were collected and immunostained for insulin and an unprocessed B-C junction of proinsulin or an unprocessed C-A junction of proinsulin (D). Representative images of n = 3, and the scale bar is 100 μm. Pancreases from Ins1^−/−Ins2^f/f mice and tamoxifen-treated Ins1^−/−Ins2^−/−Pdx1CreER mice were immunostained for insulin and the C-terminus of pro-islet amyloid polypeptide (proIAPP) or the N-terminus of proIAPP (E). Representative images of n = 2-4, insets enlarged 3×, and the scale bar is 100 μm. Western blot for IAPP and proIAPP using plasma from Ins1^−/−Ins2^f/f PdxCreER mice collected 1 week before or 8 weeks after tamoxifen treatment (F). Quantification of the relative levels of proIAPP/IAPP pre-tamoxifen vs post-tamoxifen shown on the right. Pancreas from Ins1^−/−Ins2^f/f PdxCreER mice collected 10 weeks after tamoxifen treatment was stained with periodic acid Schiff (G). Representative images of n = 3, and the scale bar is 100 μm. Electron microscopy with immunogold labeled α-prohormone convertase (H). Images of representative cells with insulin containing granules (left) or with IAPP-containing granules (right) shown. Blue circles denote granules with prohormone convertase 1/3 (PC1/3) immunoreactivity, and red circles denote granules without PC1/3 immunoreactivity. Arrows point to example immature granules. The scale bar is 500 nm, and insets are enlarged 4×. Quantification of the proportion of granules with PC1/3 immunoreactivity (I) and the proportion of PC1/3 immunoreactivity localized to secretory granules (J). Groups were compared by repeated measures 2-way analysis of variance (A-C) or paired t-test (H). Data presented as mean ± SE of the mean with individual animals in faint lines (A-C) or as individual animals with a line at the median (F and H). *P < 0.05, **P < 0.01, ***P < 0.001.

Thirteen weeks post-AAV delivery, we collected pancreases, and AAV Ins1-Ins1 had an infectivity rate of 10% to 15% (rate of C-peptide 1 immunoreactivity), and AAV Ins1-Cre had an infection rate of approximately 50% (rate of loss of C-peptide 2 immunoreactivity) [Supplemental Figure 3F (18)]. AAV Ins1-Cre–treated animals had normal proinsulin immunoreactivity (Fig. 4D), and despite decreased PC2, immunoreactivity in the core of the islet [Supplemental Figure 3G (18)] had normal processing based on observations of normal circulating proinsulin levels (Fig. 4C). Unlike PBS-treated Ins1^−/−Ins2^f/f mice, there was an increase in N- and C-terminal proIAPP immunoreactivity in the AAV Ins1-Cre–treated groups (Fig. 4E) and similar high rates of cytoplasmic proIAPP immunoreactivity in the Ins1^−/−Ins2^f/f;PdxCreER model (Fig. 4E) of adult-onset insulin-deficient diabetes [Supplemental Figure 3H (18)]. Cytoplasmic immunoreactivity for the N-terminus of proIAPP was more pronounced, potentially representing partial processing of the C-terminal fragment. Further, we note that the gross islet morphology appeared different between control mice (Ins1^−/−Ins2^f/f + PBS) and adult inducible insulin-deficient mice including both areas with apparent gaps between cells in the core of the islet and less consistent nuclear size and shape (Fig. 4E). Unlike mice treated with only AAV Ins1-Cre (wherein there were no cells with insulin and high proIAPP immunoreactivity), a subset of cells in the AAV Ins1-Cre + AAV Ins1-Ins1–treated group had both insulin and proIAPP. These cells may represent cells that had Ins2 recombination events paired with gaining Ins1 expression via AAV Ins1-Ins1 transduction, but we were unable to definitively test this hypothesis due to the lack of suitable antibodies.

In Ins2^f/f;PdxCreER mice treated with tamoxifen and with or without the Cre transgene, we assessed proIAPP processing by performing Western blots on plasma (Fig. 4F). Four bands were detected, including 1 that was parallel to an IAPP standard, 1 that was at the expected size for intact proIAPP (7.4 kDa), and 2 intermediate bands presumed to represent processing intermediates. Given that plasma was used and we aimed to assess the processing efficiency, we assessed the relative ratios of mature IAPP to proIAPP_1-70 or partially processed proIAPP_1-48 rather than comparing to a housekeeping target. Tamoxifen administration led to a greater increase in the proportion of IAPP-immunoreactive material in serum with bands at the expected size for intact proIAPP_1-70 and the intermediate proIAPP₁₋₄₈ in Cre+ animals compared to Cre− animals. We performed electron microscopy to assess granule morphology in adult inducible insulin-null β-cells. Given that cells in the core of the islets were largely devoid of granules and instead had many large vacuoles [Supplemental Figure 5 (18)], we performed periodic acid Schiff stains and observed bright pink staining, suggesting glycogen accumulation in insulin-deficient β-cell (Fig. 4G). To investigate the cause of poor proIAPP processing, we performed transmission electron microscopy on pancreases from 3 Cre+ mice treated with tamoxifen (Fig. 4H). In cells in the islet core with abundant insulin granules (presumed to be representing cell containing at least 1 nonrecombined Ins1 allele), immunogold-labeled PC1/3 antibody [directed against the N-terminus of PC1/3 and known to detect the 66 kDa form of PC1/3 (7)] produced immunoreactivity in most insulin granules, but in cells in the core of the islet with few immature granules (presumably representing IAPP granules), almost half of granules lacked PC1/3 (Fig. 4HI). Furthermore, unlike INS+ cells wherein ~80% of PC1/3 labeling localized to insulin granules, in INS− cells less than 10% of PC1/3 immunoreactivity localized to the granules (Fig. 4J), instead being observed in vacuole-like compartments and in the ER (Fig. 4H).

Discussion

Monogenic diabetes is a candidate disease for a gene therapy cure. With the availability of AAV as a clinically relevant viral vector, we investigated the viability of an AAV-based gene therapy for the Ins1^−/−Ins2^−/− mouse model of permanent neonatal diabetes. Despite delivery of the insulin gene to Ins1^−/−Ins2^−/− β-cells, AAVs carrying complimentary DNA encoding either mouse or human proinsulin could, at best, transiently reverse diabetes, likely because of low infectivity or a severe, enduring prohormone processing defect in Ins1^−/−Ins2^−/− β-cells. We first treated 2-week-old Ins1^−/−Ins2^−/− pups with AAV Ins1-INS and observed a transient remission of diabetes. We propose that circulating proinsulin was responsible for lowering blood glucose in the 1 to 2 weeks post-AAV delivery. By immunostaining pancreases, we observed impaired processing at the B/C junction in AAV Ins1-INS infected Ins1^−/−Ins2^−/− β-cells, suggesting a buildup of des-64,65 proinsulin and/or split-65,66 proinsulin, which have bioactivities ~20% that of mature insulin [Supplemental Table 3 (18)] (33, 34). Given that that these animals had elevated circulating proinsulin and presumed normal insulin sensitivity at 15 days of age (when AAV was delivered), we conclude that these proinsulin intermediates were sufficient to lower blood glucose. In the following weeks, the body weight of these mice doubled, and as animals reached 5 to 6 weeks of age, there was an expected peak in counterregulatory growth hormone during the pubertal period (35), thus increasing insulin requirements. Although the proportion of β-cells with human C-peptide immunoreactivity at the time of pancreas collection was <10% (when the mice were adults), the proportion of β-cells infected during the period of diabetic remission (~3 weeks of age) was likely much higher before rapid β-cell expansion near the time of weaning (23, 24), β-cell neogenesis (25-27), and/or degradation of viral genomes diluted the proportion of β-cells with C-peptide immunoreactivity. Retreatment of a subset of animals with extra AAV was not successful at improving infection rate or levels of circulating human C-peptide. Given that mice mount an immune response and form memory immune cells against AAVs following exposure to high-dose AAV (36), it is perhaps unsurprising that retreatment failed to effectively deliver genes to additional β-cells.

We next designed and produced an AAV carrying the native mouse Ins1 open reading frame (AAV Ins1-Ins1) and kept mice alive by islet transplantation into adulthood before delivering a high dose of AAV to target a greater population of more mature β-cells (23, 24) (15). Although limited by small sample size, we achieved higher rates of infectivity that were sufficient to yield insulin independence and near-normal blood glucose in mice after removal of the islet graft. Despite this, animals still had defective prohormone processing, leading us to investigate whether avoiding any developmental effects of insulin deficiency could prevent the observed prohormone processing defect by inducing insulin deficiency in adult mice. Additionally, exposure to hyperglycemic conditions can have dramatic effects on islets, including dedifferentiation (32), transdifferentiation (37), dysfunction (38), and reduced expression of mature β-cell factors (39). As such, it was surprising that in 2 distinct models of inducible adult insulin-deficient mice, including treating Ins1^−/−Ins2^f/f mice with a moderate dose of AAV Ins1-Cre that did not induce robust diabetes (40), we still observed impaired processing of proinsulin in AAV Ins1-Ins1–infected β-cells and impaired proIAPP processing in both AAV Ins1-Ins1–infected and neighboring uninfected insulin-deficient β-cells. Further, observations of defective proIAPP processing in tamoxifen-treated Ins1^−/−Ins2^f/fPdx1CreER mice rules out the possibility that the AAV induced defective processing in β-cells. Although low infectivity rates of AAV Ins1-Ins1 precluded us from being able to convincingly assess the importance of the impaired prohormone processing phenotype for impeding diabetes remission, observations of increased proinsulin and proIAPP immunoreactivity clearly demonstrate that β-cells that lose their endogenous insulin production develop impaired prohormone processing, even if developmental effects are avoided.

We made the discordant observation of impaired proinsulin processing despite apparently normal PC1/3, PC2, and 7B2 immunoreactivity in AAV-transduced β-cells. As 1 potential mechanism underlying defective proinsulin processing, we observed evidence for abnormal sorting of PC1/3 by electron microscopy in inducible insulin-deficient (Ins1^−/−Ins2^f/fPdxCreER) mice treated with tamoxifen. The mere expression of PC1/3 and PC2 within cells is not adequate for prohormone processing; PCs must be properly sorted into secretory granules that are adequately acidified and contain sufficient calcium for proper proenzyme maturation to gain function. Additionally, prohormones such as proinsulin must also be properly folded and sorted into secretory granules to be processed by prohormone convertases (41) but endoproteolysis of proinsulin by PC1/3 is necessary for sorting of proinsulin [and other hormones (42)] into mature secretory granules (43). Taken together, it is possible that defective PC1/3 sorting causes the proinsulin processing defect, but we cannot definitely make this conclusion. Potentially, the loss of paracrine or autocrine insulin signaling drives the defective proinsulin processing and observed association with defective PC1/3 sorting, as supported by observations that disruption of insulin receptor expression in β-cells reduced the expression of carboxypeptidase E and ultimately resulted in impaired proinsulin processing (44, 45). Alternatively, it seems more likely that since proinsulin and eventually insulin make up >50% of the protein in healthy β-cells (46), loss of insulin leads to dysregulation of intracellular trafficking (including the sorting of PC1/3), which may be irreversible, even in the presence of AAV-induced proinsulin expression. Interestingly, we also observed a buildup of glycogen in β-cells by periodic acid Schiff stain, a finding similarly observed in mouse models of hyperglycemia (47) and impaired mitochondrial metabolism (48). With evidence of impaired PC1/3 sorting in our mouse model, including into vacuole compartments, it is possible that defective sorting of PC1/3 also occurs in humans with diabetes.

Although our attempts to treat the Ins1^−/−Ins2^−/− mouse model of permanent neonatal diabetes were not fully successful, investigations into the efficacy of gene therapies to treat other models of monogenic diabetes are warranted. Most INS mutations cause diabetes not via loss of gene expression (as in the Ins1^−/−Ins2^−/− mouse model) but rather via a dominant negative mechanism. Misfolded proinsulin causes severe ER stress, and thus it is likely that such patients with mutant insulin lose β-cells after birth until they have few remaining β-cells (9, 49, 50). Without surviving β-cells, insulin gene replacement will not be a viable approach to cure insulin deficiency. Instead, assessing the viability of a gene-editing approach could be clinically relevant and informative. For example, AAV8 could be used for delivery of CRISPR-Cas9–mediated repair to Akita mice to correct their C96Y mutation and restore normal insulin production and prevent onset of diabetes (51). Alternatively, early intervention with other therapies aimed at degrading misfolded proinsulin aggregates including overexpression of chaperones (52) may preserve endogenous β-cells, thereby making insulin gene replacement therapy a more viable option. Additionally, there are patients with mutations altering insulin receptor binding affinity, with mutations that alter proinsulin sorting or processing (9), and with homozygous mutations in the intron of INS (14). These patients do not produce misfolded proinsulin and thus retain surviving insulin-deficient β-cells as evidenced by normal circulating IAPP immunoreactivity in 1 family (14). Although the current work suggests that insulin-deficient IAPP-producing human β-cells would be unable to properly produce and secrete mature insulin even after correction or replacement of the mutated insulin genes, patients with mutations that reduce the bioactivity or biosynthesis of insulin (53) or who have recessive loss of function mutations in the insulin gene (14) are potentially ideal targets for an insulin gene therapy. Based on our findings, it will be critical to assess β-cell prohormone processing for any gene therapy approaches used to treat monogenic diabetes.

Some forms of monogenic diabetes may be ideal candidates for a gene therapy approach to replace missing or damaged genes and thus cure the disease. Having suitable tools to replace or repair genes in the β-cells could be a pathway to a cure for the 1% to 6% of patients with monogenic forms of diabetes. Although the current work highlights that there likely remain major hurdles for such gene therapies for diabetes, including whether dysfunctional or surrogate β-cells are capable of properly storing and processing proinsulin into mature, fully bioactive insulin, more work is warranted. Perhaps with further progress, gene therapy approaches will be another option to helps patients live longer, less morbid lives.

Author Contributions

A.R. prepared the manuscript with contributions from R.K.B., M.M., C.B.V., and T.J.K. All authors contributed to study design, sample or data collection, and/or data analysis or interpretation. All authors approved the final version of the manuscript. T.J.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

This research was supported by funding from the JDRF and the Canadian Institutes of Health Research (CIHR). A.R. gratefully acknowledges studentship support from the CIHR (Vanier Canada Graduate Scholarship) and Vancouver Coastal Health (CIHR-UBC MD/PhD Studentship).

Disclosures

T.J.K. has received consulting fees from Fractyl Health, Inc. The other authors have nothing to disclose.

Data Availability

Some or all data sets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.