The ability to transfer immunoregulatory, cytoprotective, or antiapoptotic genes into pancreatic islet cells may allow enhanced posttransplantation survival of islet allografts and inhibition of recurrent autoimmune destruction of these cells in type 1 diabetes. However, transient transgene expression and the tendency to induce host inflammatory responses have limited previous gene delivery studies using viral transfer vectors. We demonstrate here that recombinant adeno-associated virus (rAAV) serotype 2, a vector that can overcome these limitations, effectively transduces both human and murine pancreatic islet cells with reporter genes as well as potentially important immunoregulatory cytokine genes (interleukin-4, interleukin-10), although a very high multiplicity of infection (10,000 infectious units/islet equivalent) was required. This requirement was alleviated by switching to rAAV serotype 5, which efficiently transduced islets at a multiplicity of infection of 100. Although adenovirus (Ad) coinfection was required for efficient ex vivo expression at early time points, islets transduced without Ad expressed efficiently when they were transplanted under the renal capsule and allowed to survive in vivo. The rAAV-delivered transgenes did not interfere with islet cell insulin production and were expressed in both β- and non–β-cells. We believe rAAV will provide a useful tool to deliver therapeutic genes for modulating immune responses against islet cells and markedly enhance long-term graft survival.
Attempts to use islet cell transplantation for reversing type 1 diabetes have been documented for more than two decades; however, the procedure has been largely unsuccessful (1,2). Concurrent mechanisms believed to underlie this lack of success include rejection, recurrence of anti–islet cell autoimmunity, and nonspecific islet loss because of perturbation of the graft microenvironment (e.g., inflammation, ischemia/reperfusion). A number of candidate gene products may prevent immune-mediated destruction and extend graft survival (e.g., interleukin [IL]-4, manganese superoxide dismutase, Bcl-2) (3). Furthermore, these genes may prove safer and more effective than systemic pharmacological immunosuppression because some agents are themselves potentially prodiabetogenic (e.g., cyclosporine, FK506, steroids) through imposition of increased metabolic demand. However, such studies have been limited by the lack of gene transfer vectors that are safe, efficient, and long lasting (4). Recombinant adeno-associated virus (rAAV) vectors have recently demonstrated some superiority to other viral and nonviral systems with regard to their in vivo safety, efficiency, and duration of action both in animal models and in early persistent infections in humans without known pathology and with only modest immune responses (5–10). rAAV retains these beneficial properties and therefore has the potential to be an ideal vector for in vivo gene transfer. However, previous studies have failed to demonstrate rAAV transduction of islet cells (3).
RESEARCH DESIGN AND METHODS
Pancreatic islet cells were isolated as previously described (11). Briefly, after intraductal injection of a solution containing Liberase (Boehringer-Mannheim Biochemicals, Indianapolis, IN), a whole human pancreas was subjected to mechanical shaking, and aliquots of eluate were withdrawn at various points during a 2-h period. Purification of the final islet preparation was obtained by centrifugation on discontinuous Eurocollins-Ficoll gradients followed by hand picking. Mouse islets (C57Bl/6; Jackson Research Laboratories, Bar Harbor, ME) were obtained through intraductal injection of collagenase type XI solution (Sigma, St. Louis, MO), followed by purification through repeated washings and hand picking. Islets were maintained in standard culture conditions (human-CMRL-1,066 with 5% normal human serum; mouse RPMI-1640 with 10% fetal bovine serum; 5% CO2, 24°C) until used (within 48 h). Islet purity was assessed by diphenylthiocarbazone staining, and viability was determined by staining with propidium iodide and fluorescein diacetate.
Plasmid construction, viral packaging and production, and cellular transduction.
The rAAV serotype 2 (rAAV2) vector plasmids used for these experiments are depicted diagrammatically (Fig. 3). Briefly, murine cDNAs for the cytokines IL-4 and IL-10 were cloned into the p43.2 (AAV2-ITR-containing-vector) plasmid between the XbaI site downstream from the cytomegalovirus (CMV) promoter and the XbaII site upstream from the simian virus 40 (SV40) polyadenylation signal.
rAAV2 production was performed as previously described (12). The method involves cotransfection with two plasmids by calcium phosphate coprecipitation of a permissive human cell line (HEK293). HEK293 cells were grown as monolayers (initially seeded with 6 × 108 cells) in Dulbecco’s phosphate-buffered saline (PBS) containing 5% fetal bovine serum (37°C, 5% CO2). After 18 h, the cells were transfected with different pairs of plasmids. The first nonrescuable helper plasmid (pDG) contained the rAAV2 complementing functions, rep and cap, as well as the Ad helper genes (E2a, VA RNA, and E4) required for helper function. The second vector contained a eukaryotic expression cassette and flanking inverted terminal repeats (ITRs). Transfected cells were maintained at 37°C in culture (5% CO2) for 60 h before harvest. Cells were then dissociated by treatment with EDTA, pelleted, resuspended in lysis buffer (20 mmol/l Tris, pH 8.0; 150 mmol/l NaCl; 5% deoxycholate) containing benzonase (Nycomed Pharma), and incubated for 30 min (37°C, 5% CO2). Crude lysates were clarified by centrifugation with virus-containing supernatant purified by iodixanol density gradient centrifugation, followed by heparin affinity chromatography and concentration. The purity of preparations was determined by subjecting the sample to silver-stained SDS-PAGE. Infectious center assays were used to determine the rAAV titer, and dot blot assays were performed to quantify the titer of the rAAV physical particles and particle-to-infectivity ratio (12). Intact islets, maintained as previously indicated at concentrations of 0.2 × 103 to 1 × 103, were transduced at a multiplicity of infection (MOI) of 10 to 10,000 infectious units (i.u.) per cell of the appropriate rAAV construct. Islet equivalents were determined for all pancreatic isolations; an estimate of 2,000 cells per islet equivalent was used for all calculations. Specifically, an islet equivalent (i.e., a combination measure of size and number) was defined as an islet that was spherical in shape and 150 μm in diameter; an appropriate algorithm was used to calculate the islet equivalent number. Using this islet equivalent value, we used 2 × 104 to 2 × 107 i.u. of rAAV per islet equivalent, which equated to an MOI of 10–10,000 i.u. per islet cell. For studies using adenovirus (Ad) as a helper virus, islet cells were treated with adenovirus 5 (Ad5) at an MOI of 5 for 2 h (37°C, 5% CO2) before coinfection with rAAV.
The comparison of rAAV2 and rAAV serotype 5 (rAAV5) vectors was performed using an expression cassette consisting of a Rous sarcoma virus (RSV) long-terminal repeat promoter and a nuclear-targeted β-galactosidase (nlacZ) transgene, flanked by either rAAV2-ITRs (13) or rAAV5-ITRs (14). The rAAV2-nlacZ construct was packaged as described above, by cotransfection of the vector plasmid with the 5RepCapB helper plasmid (14) into Ad5-infected cos cells and purified by CsCl ultracentrifugation.
Measurement of cytokine and insulin production.
Microtiter plates (Immulon) were coated with 50 μl of a 1:250 dilution of anti-mouse IL-4 or IL-10 (#265113E, #26571E; Pharmingen, San Diego, CA) in 0.1 mol/l sodium bicarbonate buffer (overnight, 4°C). After washing and appropriate blocking (with 10% fetal bovine serum in PBS-Tween 20, 1 h at 24°C), standards for IL-4 or IL-10 and tissue culture medium samples were incubated in the plate at 24°C for 1 h. After washing, a second antibody (1:250 dilution of horseradish peroxidase–conjugated anti-mouse IL-4, #26517E, or 1:250 dilution of biotylated anti-mouse IL-10, #26572E, with streptavidin–horseradish peroxidase conjugate) was reacted with the captured antigen at 24°C for 1 h. After extensive washing, detection was performed using a third incubation with O-phenylenediamine (Pharmingen) solution detection and measurement of the absorbance at 490 nm. For the detection of insulin, supernatants were processed and hormone secretion was quantitated using commercial kits (Mercodia, Minneapolis, MN). Data are expressed as means ± SE.
For insulin immunolocalization, intact human islets were ethanol fixed and rehydrated through repeated washings in solutions containing decreasing ethanol concentrations (99, 95, 70, and 0%; 30 s, 24°C). After being washed in PBS, islets were incubated for 1 h at 24°C with 0.5 μg/ml guinea pig monoclonal anti-insulin antibody (Dako). Primary antibody was detected after standard washing and blocking steps, including incubation (1 h, 24°C) with biotinylated goat anti–guinea pig antibody coupled thereafter with streptavidin-RPE-Cy5 (Dako). Fluorescent microscopy was performed using a Zeiss Axioplat unit, and confocal microscopy was performed using a Bio-Rad 1024ES laser scanning confocal system attached to an Olympus SX70 inverted microscope.
RESULTS AND DISCUSSION
rAAV binds to cells via a heparan sulfate proteoglycan receptor. After it has been attached, its entry is dependent on the presence of a coreceptor, which may consist of either the fibroblast growth factor receptor or the αv-β5 integrin molecule (15,16). To readdress the question of whether islet cells were permissive for rAAV vectors, we performed a series of transduction experiments with purified human islets. These initial experiments used both the UF5 rAAV-CMV–green fluorescent protein (GFP) vector and the UF11 rAAV–CMV/β-actin (CB) hybrid promoter–GFP vector (12). Batches of 1 × 103 intact human islets were infected at an MOI of 10 to 10,000 i.u. per islet equivalent. To enhance scientific interpretation of these short-term in vitro experiments, islets were coinfected with Ad5 at an MOI of 5. This coinfection procedure results in an acceleration of leading strand synthesis (7,13) but is not an absolute requirement for rAAV transgene production. Standard fluorescent as well as confocal microscopy revealed that GFP expression was quite efficient (i.e., >40% GFP-positive cells by computer-aided morphologic assessment) in human islets within 48 h of infection under these conditions (Figs. 1A and B). Interestingly, transduction was much less efficient (<1% GFP-positive cells) at an MOI of 1,000, was indistinguishable from control vector at an MOI of 100 or less, and was of similar efficacy (at equivalent MOI) using either CMV- or CB promoter–based systems. We believe our use of a high MOI, combined with benefits afforded through recent improvements in rAAV purification methods (12), led to this novel finding of islet cell rAAV transduction, which was not observed in previous studies (3).
Although the successful transduction of human islets represents an important finding in terms of affording the feasibility for future clinical intervention in humans, identifying rAAV transduction in rodent islets is considered crucial for investigations exploring the effects of therapeutic transgene expression in experimental transplantation models. To address the issue of species specificity and permissiveness for islet rAAV transduction, identical studies (both in terms of MOI titration and testing of CMV and CB promoters) were extended to intact islet cells obtained from mice. Similar to human islets, successful transduction (as demonstrated by rAAV-GFP expression [Figs. 1C and D]) was achieved with titration and promoter efficiencies overlapping those observed with their human cellular counterparts.
Another key issue for therapeutic efficacy concerns the distribution of rAAV-GFP expression within an islet in combination with the question of whether rAAV transduction of β-cells occurs. To address these issues, a series of intact islets were transduced with rAAV-GFP and subjected to confocal imaging (single slice of a transduced human islet [Fig. 2A]), a procedure that revealed homogenous GFP expression throughout the islet. To identify whether β-cells were capable of transduction, cytocentrifuged preparations of these rAAV-GFP–transduced human islet cells were immunostained with an RPE-Cy5–conjugated anti-insulin antibody (Fig. 2B). Fluorescence microscopy revealed colocalization of staining in β-cells (red anti-insulin stain, green rAAV-GFP fluorescence), indicating that this cell type had been effectively transduced (Fig. 2B, inset).
Depending on the mode of administration (i.e., systemic versus local), treatment with the immunoregulatory cytokines IL-4 and IL-10 can inhibit the recurrence of type 1 diabetes (alloimmune and/or autoimmune) in mice receiving islet transplants; IL-4 seems to inhibit disease-causing lymphocytes and IL-10 seems to limit the activation of potentially diabetogenic CD8+ T-cells (17–19). However, use of cytokines for initiation of immune deviation systemically would currently be limited because of the need for repeated administration, because of their relatively short half-life, and local production, which is dependent on the availability of suitable targeted gene delivery systems (20,21). A modification of islet cells toward production of these anti-inflammatory cytokines, achievable by rAAV gene transfer, could be exploited for developing novel immunointervention protocols for type 1 diabetes.
To address this potential strategy, human pancreatic islets were transduced with rAAV vectors containing the cytokines IL-4 and IL-10. Specifically, experiments were performed wherein at an MOI of 10,000, intact human islets were transduced with rAAV-CMV-IL-4 or rAAV-CMV-IL-10 (Fig. 3A). At 48 h, both IL-4 and IL-10 were readily detectable from treated islets (1.32 ± 0.62 and 2.23 ± 0.34 ng/ml per 0.2 × 103 islets for IL-4 and IL-10, respectively; Fig. 3B), whereas levels of these cytokines were not detectable from Ad-infected islets transduced with GFP or irrelevant rAAV control vector preparations. These data demonstrated successful rAAV-mediated islet cell transduction with a potentially therapeutic secreted protein. It was also of interest to observe whether the transduction of islets with rAAV interfered with β-cell metabolic function. To address this issue, sets of 0.4 × 102 intact human islets (in triplicate) were used as a control or transduced with the UF5 rAAV-CMV-IL-10 vector (MOI 10,000 i.u.) plus Ad5 (MOI 5 i.u.), UF5 rAAV-CMV-IL-10 vector (MOI 10,000 i.u.) alone, or Ad5 (MOI 5 i.u.) alone. The islets were maintained for 48 h under basal (5 mmol/l glucose) or stimulated (20 mmol/l glucose) conditions; medium samples were withdrawn at 0, 2, 12, 24, and 48 h for analysis of insulin production. Although conditions of elevated glucose imparted a two- to threefold increase in insulin release, no differences (analysis of variance, NS) in insulin release were detected between the control and the groups of transduced islets within the two conditions of glucose stimulation (Fig. 3C). Repeat experiments performed with islets from which samples were collected at 24-h intervals over a 1-week interval provided similar results, in that no differences in insulin release were identified. In general, neither wild-type AAV nor rAAV has ever been shown to induce apoptosis. However, we formally excluded this possibility by measuring free lactate dehydrogenase in the supernatant media after infection (data not shown). The lack of a lactate dehydrogenase increase indicated that islets remained viable throughout the course of the experiment. In addition, a fine metabolic assessment of islet cell function using a static glucose stimulation assay has been performed; studies further support the contention that rAAV-transduced islets are not impaired in terms of their glucose responsiveness. Specifically, islets respond to low to high-to-low (i.e., 1.67 to 16.7 to 1.67 mmol/l) sequential incubations with appropriate insulin secretion levels (E. Fenvez, L. Inverardi, unpublished observations).
Although these data were very promising, the very high MOI of rAAV required could present a significant limitation to clinical gene therapy. We hypothesized that a scarcity of receptors for AAV2 on islet cells could account for this. To test this hypothesis, we compared the transduction efficiency of rAAV2 at a lower MOI (100 i.u.) with that of rAAV serotype 5, which uses a different attachment receptor (14,22,23). Cultures containing 1 × 105 Ad-infected murine islet cells were transduced with 1 × 107 infectious units of either rAAV2-RSV promoter-driven nuclear-localized lacZ vector or 1 × 107 infectious units of rAAV5-RSV-nlacZ. These intact islet cells were cultured for 48 h and then stained with X-gal for 4 h before imaging. As previously shown, rAAV2 was ineffective at this dose, whereas rAAV5 resulted in abundant lacZ-positive nuclei (Fig. 4). This finding was consistent with our hypothesis that AAV2 receptors are limiting and indicated a possible role for rAAV5-based vectors in future studies.
Finally, it will be necessary for rAAV islet cell gene therapy to be effective in the absence of Ad augmentation and to be stable after transplantation. We had hypothesized, based on earlier studies, that conversion of rAAV from ss-DNA to ds-DNA form in the absence of Ad would require at least 7–10 days (13). Furthermore, studies with a related rAAV2-α1-antitrypsin vector showed that rAAV expression in human islets transduced without Ad was measurable at 2.5-fold above background by day 3 and peaked at 3.5-fold above background by days 7–8 (data not shown). To more formally test this, we transplanted a bolus of 1,000 pancreatic islets transduced with the UF11 vector in the absence of Ad under the renal capsule of syngeneic C57Bl/six mice. Mice were killed 2 weeks later, and the site of the graft was analyzed by epifluorescence. Transduced islets showed bright green native GFP fluorescence (Fig. 5B), whereas the surrounding kidney parenchyma and control kidney showed very little background autofluorescence (Fig. 5A). The efficiency of GFP expression at 2 weeks without Ad was greater than that seen with Ad at the earlier time points. As with previous in vivo studies with rAAV (8,10,24,25), a carefully controlled histopathological examination of parallel hematoxylin-eosin–stained sections showed no evidence of inflammation or cellular infiltration within or near the implanted islets.
Despite advances in disease management, life span is shortened by one-third in patients in whom type 1 diabetes develops before age 30 years, and many patients develop significant microvascular and macrovascular complications (26). These complications are the reason diabetes is recognized as a leading cause of blindness (retinopathy), heart disease, peripheral vascular disease, renal failure, and impotence. Because there are no known innovations that seem likely to alter this situation soon, a novel approach to preventing pancreatic β-cell destruction after islet cell transplantation is appealing.
The findings of this study demonstrate feasibility for a novel and exciting strategy that could dramatically improve islet cell transplantation through genetic engineering of islets. Our approach was to devise an efficacious gene delivery system that would allow for testing the question of whether immunoprotection could be afforded through local cellular production of immunomodulatory cytokines. This approach is very topical because the role of immunomodulatory cytokines, in terms of allograft and xenograft rejection, and recurrent autoimmunity are subjects of current interest and controversy (27). However, it should be noted that the introduction of cytoprotective molecules into islets using rAAV vectors will not be limited to cytokines. A series of recent studies has indicated pivotal roles for both antioxidants (e.g., heme-oxygenase-1, manganese superoxide dismutase) and agents capable of interrupting apoptotic pathways (e.g., Bcl-2, survivin) in prolonging graft survival after transplantation. Hence, such molecules form obvious candidates for future investigations using rAAV as a vector for gene delivery into islets. Furthermore, the therapeutic benefits of rAAV delivery of cytoprotective molecules may also be expanded to other systems of solid organ transplantation in which ex vivo manipulation of grafts before engraftment is possible. It will be key for future studies to demonstrate unequivocally that rAAV transduction in the absence of Ad leads to long-term expression in islet cells in vivo, a factor that would greatly bolster arguments affording this method of therapy as a safer approach. The unique properties of rAAV versus other gene therapy delivery approaches provide an ideal treatment modality for delivery of transgenes for long-term expression in islets before transplantation and one that could markedly enhance the efficacy of this clinical procedure.
This work was supported by grants from the National Institutes of Health (DK-51809, HL-59412), the Juvenile Diabetes Foundation International, and endowments from the S. Family/American Diabetes Association Chair and the Powell Gene Therapy Center.
The authors acknowledge the advice and technical assistance of Drs. Camillo Ricordi and Robert Kotin. The islets used for this study were provided by the Juvenile Diabetes Foundation Islet Isolation Facility at the University of Miami. Murine cDNAs for the cytokines IL-4 and IL-10 were kindly provided by Dr. Nora Sarvetnick, Scripps Institute, La Jolla, CA.
Address correspondence and reprint requests to Dr. Mark Atkinson, University of Florida, Department of Pathology, Box 100275 JHMHC, 1600 SW Archer Road, Gainesville, FL 32610. E-mail: email@example.com.
Received for publication 23 June 2000 and accepted in revised form 1 November 2000.