Nonobese diabetic (NOD) mice develop insulitis and diabetes through an autoimmune process. Since TGF-β1 down-regulates many immune responses, we hypothesized that TGF-β1 could prevent disease in NOD mice and that there would be several advantages to cytokine delivery by a somatic gene therapy approach. We opted for i.m. injection of a naked plasmid DNA expression vector encoding murine TGF-β1 (pCMV-TGF-β1). Treatment with pCMV-TGF-β1 resulted in the retention and expression of the vector in muscle cells, associated with a considerable elevation in the plasma levels of TGF-β1, that was not observed in control vector-treated mice. The levels of TGF-β1 produced were sufficient to exert immunosuppressive effects. Delayed-type hypersensitivity responses were suppressed, and autoimmunity-prone NOD mice were protected from insulitis and diabetes in models of cyclophosphamide-accelerated and natural course disease. In pCMV-TGF-β1-treated mice, pancreatic IL-12 and IFN-γ mRNA expression was depressed, and the ratio of IFN-γ to IL-4 mRNA was decreased, as determined by semiquantitative reverse-transcription PCR. In contrast, NOD mice injected with a vector encoding the proinflammatory cytokine IFN-γ developed diabetes earlier. Intramuscular administration of cytokine-encoding plasmid vectors proved to be an effective method of cytokine delivery in these mice, and altered autoimmune disease expression.
The nonobese diabetic (NOD)3 mouse develops diabetes spontaneously, through an autoimmune process. This disease shares many features with human insulin-dependent diabetes mellitus (IDDM or type 1 diabetes) (1). There is a T cell- and macrophage-dependent progressive infiltration of islets of Langerhans (termed insulitis), with destruction of insulin-producing β islet cells, occurring over a period of weeks or months. Strong experimental evidence suggests that proinflammatory cytokines produced by Th1 cells and macrophages play an important role in the development of these lesions (2, 3). Therefore, suppressing the activity of these cells is likely to be therapeutically effective.
In this study, we focused on the potential immunosuppressive role of TGF-β1 in autoimmune murine diabetes. TGF-β1 is a pleiotropic cytokine with multiple antiinflammatory effects. It suppresses the activity of T cells, macrophages, NK cells, and B cells, and inhibits the expression of many proinflammatory cytokines such as IFN-γ, TNF-α, IL-1, and IL-2 (4). We hypothesized that TGF-β1 could prevent IDDM in NOD mice, and that there would be several advantages to delivering this cytokine by a somatic gene therapy approach.
It has been shown that i.m. injection of naked plasmid expression vectors results in the cellular uptake of the plasmid DNA, which is maintained episomally for prolonged periods of time within skeletal muscle cells (5). In this study, we demonstrate that i.m. injection of a vector encoding mouse TGF-β1 cDNA, i.e., pCMV-TGF-β1, results in uptake, retention, and expression of this vector. There is detectable vector-derived TGF-β1 mRNA in skeletal muscle cells, as well as increased levels of TGF-β1 in the plasma of treated mice. Administration of pCMV-TGF-β1 was effective at suppressing a DTH response, and at protecting NOD mice from insulitis and diabetes. There is a decreased expression of IL-12 and IFN-γ mRNA in the pancreas of protected mice. In contrast, administration of an IFN-γ-encoding vector accelerated disease.
Materials and Methods
Female NOD mice (8–10 wk) were purchased from Taconic Farms (Germantown, NY), and female BALB/c mice (substrain AnNCrlBR, 4–6 wk) were purchased from Charles River Canada (St-Constant, Quebec, Canada). Mice were kept in a pathogen-free facility. These NOD mice exhibit mild insulitis as early as 4 wk of age and become diabetic starting at about 12 wk of age.
Expression vector construction and TGF-β1 assays
The mouse TGF-β1 (mTGF-β1) cDNA was produced by RT-PCR from Con A-stimulated BALB/c splenocytes and cloned into compatible enzyme restriction sites of pCI-neo (Promega, Madison, WI) to generate pCMV-TGF-β1. The mTGF-β1 cDNA is under the transcriptional control of a CMV immediate-early enhancer/promoter, and downstream of a chimeric intron. pCMV-TGF-β1 encodes the latent form of mTGF-β1, and expression was confirmed using a TGF-β1 ELISA (R&D Systems, Minneapolis, MN) in supernatants collected from transiently transfected COS-7 cells. The TGF-β1 ELISA only detects active TGF-β1. The latent TGF-β1 was activated to its biologically active form by acidification for 10 min, and its bioactivity was confirmed by the CCL64 mink cell line proliferation assay. Mouse IFN-γ cDNA was subcloned into pCI-neo, as described above, to generate pCMV-IFN-γ. pCI-neo, henceforth referred to as pCMV-null, was used as control vector in all of the experiments.
Plasmid DNA preparation
Large-scale plasmid DNA preparations were produced by the alkaline lysis method using a Qiagen giga kit (Qiagen, Santa Clarita, CA). All plasmid preparations for i.m. injections were resuspended in sterile 0.85% saline. Spectrophotometric analysis revealed 260/280 nm ratios ≥1.80. Purity of DNA preparations and conformations was confirmed on a 1% agarose gel.
Determination of plasma TGF-β1 levels
To minimize the activation of platelets and subsequent release of endogenous TGF-β1, platelet-poor plasma was obtained as follows: whole blood was mixed with a 1.5% EDTA solution, mixed thoroughly, and put on ice immediately after blood collection. The blood/EDTA mixture was then layered gently on 20% sucrose and centrifuged for 30 min at 12,000 × g. The upper two-thirds of the upper phase (platelet-poor plasma fraction) were collected without disturbing the interface. Once separated, the platelet-poor plasma was frozen at −80°C until assayed for TGF-β1 by ELISA. Statistical analysis was performed by Student’s t test.
Induction and measurement of DTH reaction in mice
On day 1, BALB/c mice were immunized with 200 μg of OVA in CFA, or with CFA alone. On day 6, mice were injected with OVA in PBS in the right footpad. DTH responses were measured, with calipers, as the increase in footpad thickness 24, 36, and 48 h after OVA Ag recall immunization. Statistical analysis was performed with the Student’s t test.
Intramuscular injection of plasmid DNA
Intramuscular injections of plasmid DNA were done as described (6). Briefly, mice were anesthetized by i.p. injection with xylazine (10 mg/kg) and ketamine (200 mg/kg). The rectus femoris (RF) and tibialis anterior (TA) muscles of each mouse were injected with a 0.5-cc sterile 29G1/2 insulin syringe, fitted with a plastic collar to limit needle penetration to 2 mm. Mice received 100 μg of pCMV-TGF-β1 in 50 μl of sterile saline in each RF or TA muscle, for a total of 400 μg of plasmid DNA in each treatment session, unless stated otherwise. Control mice received equivalent amounts of pCMV-null control vector in each muscle group. A vector encoding mouse IFN-γ, pCMV-IFN-γ, was administered in a similar fashion.
Induction and diagnosis of diabetes
Cyclophosphamide (CYP; Sigma, St. Louis, MO), which accelerates the onset of diabetes in NOD mice (7), was administered i.p. twice, 14 days apart, at a dose of 200 mg/kg in PBS to 8- to 10-wk-old female NOD mice. Diabetes was diagnosed by regular urinary glucose analysis and confirmed by blood glucose determination. Mice were considered diabetic when sequential blood glucose measurements were shown to be equal to, or above 16.7 mmol/L (300 mg/dl), as determined by the Accu-ChekIII glucometer (Boehringer Mannheim, Indianapolis, IN).
Statistical analysis was done using the SAS software (version 6.12) for Windows 95. The incidence of diabetes was plotted using the Kaplan-Meier method (nonparametric cumulated survival plot). The statistical comparison between the curves obtained was performed using the Wilcoxon log rank test.
The pancreas was excised immediately after CO2 asphyxiation and fixed in 10% buffered Formalin, and hematoxylin and eosin-stained histologic slides were prepared. Insulitis was graded as follows: grade 0, normal islet totally free of any periislet mononuclear cells; grade 1, focal periislet lymphocytic infiltration <25% of islet circumference; grade 2, periislet lymphocytic infiltration >25% of islet circumference; grade 3, mild insulitis, intraislet infiltration with good retention of islet cell morphology; grade 4, severe insulitis with significant destruction of β-islet cells. Three randomly obtained levels of pancreas were analyzed in double-blind fashion by two observers. Statistical analysis was performed with the χ2 test.
PCR and RT-PCR analysis
Mice were killed and their TA muscles were excised, immediately frozen in liquid nitrogen, and stored at −80°C. Total genomic DNA was isolated from thawed muscle specimens, as described (8). The PCR reactions were performed in a 50 μl reaction vol containing 2.5 μl genomic DNA, 10 mM Tris-HC1, pH 8, 50 mM KC1, 2 mM dNTP, 5 mM MgC12, 2 μM of each primer, and 1.5 U Taq DNA polymerase. The primer sequences amplifying vector-encoded TGF-β1 were 5′-AGAGAAGAACTGCTGTGTGCGGCAG-3′ (sense) and 5′-CGCTTCCCTTTAGTGAGGGTTAATG-3′ (antisense). The TGF-β1 primer set amplified a TGF-β1 product from pCMV-TGF-β1 DNA, or cDNA derived from that vector, but not from genomic DNA or endogenous TGF-β1 cDNA. PCR cycling conditions were as follows: one cycle at 94°C; 40 cycles at 94°C for 1 min, at 55°C for 2 min, and 72°C for 2 min; and one final extension cycle at 72°C for 10 min. The PCR amplifiers were analyzed on a 1.5% agarose gel containing 0.5 μg/ml ethidium bromide. The TGF-β1 mRNA in treated muscle was detected by RT-PCR. Total RNA was extracted from entire TA muscles, as described (8), and reverse transcribed with the Superscript preamplification system (Life Technologies, Gaithersburg, MD). A total of 2 μl of the reverse-transcription reaction was used for PCR amplification using the above-mentioned vector-specific primers for pCMV-TGF-β1, or G3PDH primers (sense, 5′-TGAAGGTCGGTGTGAACGGATTTGGC-3′, and antisense, 5′-CATGTAGGCCATGAGGTCCACCAC-3′). Optimal amplification conditions were conducted for 40 cycles and amplifiers were labeled with [α-32P]dCTP (5 μCi) (ICN, Mississauga, ON). The PCR products were analyzed on 1.5% agarose gel containing 0.5 μg/ml ethidium bromide, transferred onto Hybond-N+ nylon membrane (Amersham Canada, Oakville, ON), and either exposed to autoradiographic film or subjected to phosphor imager analysis.
For intrapancreatic cytokine mRNA analysis, total RNA was isolated from snap-frozen pancreas and quantification of specific cytokine gene expression was performed by RT-PCR, as described above. For this purpose, specific primers were used in PCR for IL-12 p40 (9), IL-4 (10), IFN-γ (10), and β-actin (11). Analysis of cytokine PCR products was performed as previously described (11). Briefly, PCR reactions were terminated in the linear portion of the amplification reaction (which extended up to 35 cycles). The 32P-labeled PCR products were analyzed on a 2% agarose gel, transferred onto a nylon membrane, and subjected to phosphor imager analysis. Semiquantitative mRNA analysis was performed by calculating relative quantities of RT-PCR signals for each cytokine, normalized to the β-actin signal of each sample. The ratio of IFN-γ to IL-4 mRNA (for a mouse) was derived after normalization of these cytokines with their β-actin signal, as described above: IFN-γ/IL-4 ratio = [(IFN-γ/β-acton ratio)/(IL-4/β-actin ratio)].
In all of these experiments, specificity of PCR products was confirmed by restriction enzyme analysis. Statistical analysis was performed with Student’s t test.
Detection of pCMV-TGF-β1 plasmid DNA and TGF-β1 gene expression in skeletal muscle
To determine whether plasmid DNA is successfully retained and expressed by treated muscle tissue, each TA muscle was injected with 100 μg of pCMV-TGF-β1, followed by total DNA or RNA extraction, and PCR or RT-PCR analysis, respectively. We found that 14 days following i.m. injection of pCMV-TGF-β1, a 348-bp amplifier, specific for this vector, could readily be detected by PCR analysis from all treated muscle samples (Fig. 1,A), and was undetectable in null-vector-treated mice (Fig. 1,B). Furthermore, RT-PCR analysis of RNA samples extracted from these muscles revealed the presence of vector-derived TGF-β1 transcripts (Fig. 1,C), which were absent in pCMV-null-treated mice (Fig. 1 D). No product was identified from RNA preparations when reverse transcription was omitted, and the PCR primers could not amplify endogenous TGF-β1 from either genomic DNA or cDNA (not shown).
Increased TGF-β1 plasma levels
To establish the optimal dose of plasmid DNA, increasing amounts of pCMV-TGF-β1 were injected at different muscle sites, and TGF-β1 levels in platelet-poor plasma were measured by TGF-β1 ELISA, 75 h postinjection. We found that vector expression, as determined by plasma TGF-β1 levels, increases with the number of injection sites and amount of DNA injected per site, up to 100 μg of DNA per injection site (data not shown). To determine whether the elevation of TGF-β1 levels was sustained, female NOD mice were injected with a total of 200 μg of plasmid DNA (100 μg in each TA muscle). TGF-β1 levels of 9.2 ± 0.4 ng/ml were observed at 36 h postinjection, compared with 2.6 ± 0.4 ng/ml in pCMV-null-treated mice (Fig. 2). These levels decreased to a mean plasma level of 4.3 ± 0.3 ng/ml on day 14 (p < 0.05 at 36 h and 14 days versus null-vector-treated mice). At day 35, TGF-β1 was still slightly higher in the pCMV-TGF-β1-treated mice, but this was no longer statistically significant. TGF-β1 plasma levels in age-matched, untreated NOD mice were 2.3 ± 0.5 ng/ml, a value statistically different from the pCMV-TGF-β1-treated mice at 36 h and day 14 (p < 0.05), but not statistically different from pCMV-null treated mice at any time point. In all cases, active TGF-β1 was either absent or present at undetectable levels in the plasma of treated mice (not shown).
Suppression of DTH responses
To examine whether i.m. injections of plasmid DNA result in sufficient production of TGF-β1 to exert biologic effects, we used a DTH model. DTH responses were assessed by footpad swelling 24 to 48 h following recall immunization with OVA in female BALB/c mice. We found that pCMV-TGF-β1-treated mice had significantly suppressed DTH responses (Fig. 3). These mice had a sevenfold reduction in DTH responses observed at 36 h, versus pCMV-null-treated mice, consistent with the in vivo production and activation of TGF-β1.
Prevention of CYP-induced diabetes by pCMV-TGF-β1 administration
To determine whether administration of pCMV-TGF-β1 could prevent autoimmune disease expression, we administered the TGF-β1 plasmid expression vector to adult NOD female mice. The autoimmune basis of this disease is well established (12). To accelerate disease expression, female NOD mice were injected with CYP, as described (13). CYP has been reported to accelerate disease in NOD mice by enhancing intraislet production of IFN-γ and other inflammatory mediators (14). Briefly, CYP was administered i.p., on days 3 and 16, at a dose of 200 mg/kg. Animals received a total of 100 μg of pCMV-TGF-β1 in sterile saline in each RF and TA muscle for a total of 400 μg of plasmid DNA, 48 h before each CYP injection. Control animals received equivalent amounts of pCMV-null control vector in each muscle. The incidence of IDDM was evaluated by sequential measurements of blood glucose levels.
Administration of pCMV-TGF-β1 significantly reduced the incidence of diabetes in NOD female mice (Fig. 4 A). The first case of diabetes occurred 14 days later in the pCMV-TGF-β1 group, compared with control group. By day 32 of the experiment, the incidence of diabetes was four times higher in pCMV-null-treated mice compared with mice receiving pCMV-TGF-β1 (p < 0.001). In diabetic mice, the hyperglycemia (≥300 mg/dl) was maintained 2 wk or more following onset of disease (data not shown).
Acceleration of CYP-induced diabetes by pCMV-IFN-γ administration
Diabetes occurred earlier in pCMV-IFN-γ-treated mice compared with control mice (Fig. 4 A, p = 0.05, versus null-vector-treated mice). The pCMV-mIFN-γ-treated mice had increased serum levels of IFN-γ (up to 200 pg/ml, as determined by ELISA), while this cytokine was undetectable in the serum of control mice (data not shown).
Protective effect of pCMV-TGF-β1 treatment on the natural course of diabetes
To determine an effect on the natural course of disease, we injected female NOD mice (9–11 wk) with pCMV-TGF-β1 in the absence of CYP (Fig. 4 B). Under these conditions, the incidence of diabetes was again significantly reduced, to approximately 50% of control values over the course of several weeks (p < 0.002). Thus, TGF-β1-mediated protection is not a feature unique to the CYP-accelerated diabetes model.
Protection from insulitis in pCMV-TGF-β1-treated NOD mice
Insulitis was graded histologically based on mononuclear cell infiltration of pancreatic islets. In Figure 5, we compare nondiabetic mice necropsied before the development of overt diabetes. In CYP-accelerated disease (Fig. 5,A), administration of pCMV-TGF-β1 reduced the mean insulitis score from 2.91 in control mice to 1.92 in treated mice (p < 0.001). Protection was also observed in natural course disease (Fig. 5,B), in which the mean insulitis score was 3.44 in control mice and 2.25 in treated mice (p < 0.01). Thus, pCMV-TGF-β1 treatment induced a shift to lower grade lesions, including increased number of normal islets (Fig. 5, A and B). This shift was more marked in CYP-induced disease than in natural course disease, but was highly significant in both cases.
Reduced pancreatic IL-12 and IFN-γ mRNA expression in pCMV-TGF-β1-treated NOD mice
Analysis of intrapancreatic cytokine expression patterns in CYP-treated mice showed marked differences between the prediabetic pCMV-TGF-β1-treated and control mice. Compared with pCMV-null control mice, pCMV-TGF-β1-treated mice had lower levels of both IFN-γ (p = 0.002) and IL-12 (p < 0.001) mRNA (Fig. 6,A). Mean IL-4 mRNA levels were not altered significantly by pCMV-TGF-β1 administration (a slight decrease versus control mice was not statistically significant) (Fig. 6,A). Consequently, there was a considerable decline of the ratio of IFN-γ to IL-4 mRNA levels (p < 0.05) (Fig. 6 B).
Cytokine therapy can influence the outcome of autoimmune diseases, by altering either Th1/Th2 balance, macrophage activity, or inflammatory versus suppressive cytokine production (15, 16). In this study, we focused on the potential immunosuppressive role of TGF-β1 in autoimmune murine diabetes. TGF-β1 is a multifunctional cytokine with numerous antiinflammatory effects (reviewed in Refs. 4 and 17). Therefore, we hypothesized that TGF-β1 could prevent IDDM in NOD mice, and we opted for delivery of this cytokine by a somatic gene therapy approach, consisting of i.m. injection of TGF-β1-encoding plasmid expression vector. Direct i.m. administration of naked plasmid DNA has been shown to be an effective route of gene delivery in vivo (5, 6, 18, 19, 20), and our objective was to use skeletal muscle as a source of TGF-β1.
In this study, we demonstrate that i.m. injection of plasmid DNA-encoding latent TGF-β1 (pCMV-TGF-β1) results in uptake, retention, and expression of this vector by muscle cells. As demonstrated by RT-PCR, there is detectable vector-derived TGF-β1 mRNA in skeletal muscle cells, which is not seen in null-vector-treated muscles. Moreover, plasma samples collected from pCMV-TGF-β1-treated mice show significantly elevated levels of TGF-β1 for well over 2 wk. Over that time period, the TGF-β1 levels were two- to fourfold higher than those of control mice.
We found that DTH responses to OVA are markedly suppressed in pCMV-TGF-β1-treated mice. Thus, it is apparent that sufficient quantities of TGF-β1 are produced to exert an immunosuppressive effect. Moreover, the production of TGF-β1 protects NOD mice from an autoimmune disease.
NOD mice spontaneously develop an autoimmune form of insulitis, with destruction of β-islet cells (reviewed in 12 . Insulitis is the precursor lesion, leading to diabetes when it is sufficiently severe. This disease is clearly T cell and macrophage dependent, although the mechanisms of islet cell destruction is not fully elucidated. Islet cells may be killed by infiltrating CTLs, NK cells, and macrophages (12). Proinflammatory cytokines (IL-1, IFN-γ, TNF-α), as well as nitric oxide (NO), are toxic to islet cells (14, 21). Th1-dependent immunity appears to be a key pathogenic factor, since, for example, administration of IL-12 enhances Th1 reactivity to islet cell Ags and rapidly induces diabetes (22). IL-12 and IFN-γ are produced locally in inflamed islets, and insulitis can be induced by adoptive transfer of islet-reactive Th1 clones (23). The administration of CYP to NOD mice stimulates intraislet production of IFN-γ and other inflammatory mediators, and results in rapid onset of diabetes (14).
We found that administration of pCMV-TGF-β1 considerably reduced incidence of diabetes in NOD mice. In CYP-induced diabetes, there was a fourfold reduction in incidence (day 32 of the experiment). In non-CYP-treated mice (natural course), treatment delayed the appearance of the first case of diabetes, and subsequently reduced the incidence of disease by approximately 50% over the course of several weeks. In the latter mice, treatment was begun at 9 to 11 wk, i.e., at a time when insulitis is already apparent, suggesting that an ongoing autoimmune response was suppressed.
pCMV-TGF-β1-treated mice had low insulitis scores, with markedly increased numbers of normal, or islets with mononuclear cell infiltrates limited to the periislet area (periinsulitis). In TGF-β1-treated mice following CYP administration, analysis of pancreatic cytokine mRNA expression by semiquantitative RT-PCR revealed depressed IL-12 and IFN-γ levels. IL-4 levels were not significantly altered. Consequently, the ratio of IFN-γ to IL-4 mRNA was reduced by TGF-β1 plasmid DNA therapy. Protection from insulitis was somewhat greater in CYP-accelerated than in natural course disease, but was statistically significant in both cases. Taken together, our results reveal that TGF-β1 plasmid therapy had an antiinflammatory or immunosuppressive effect. As expected, mice that were protected from insulitis were also protected from diabetes.
The immunoregulatory influence of TGF-β1 has been studied in many in vivo models, including experimental allergic encephalomyelitis (24), collagen-induced arthritis (25), and allograft rejections (26). In all of these instances, microgram amounts of TGF-β1 were administered, over a period of a few days, to achieve a significant suppression of autoimmune responses. Since we could suppress DTH and prevent autoimmune disease expression with i.m. injections of pCMV-β1, it appears that we are achieving results equivalent to administration of microgram quantities of TGF-β1 protein.
TGF-β1 may act at multiple levels to block inflammatory reactions and/or prevent autoimmune disease. The potent immunosuppressive effects of TGF-β1 were clearly demonstrated in the TGF-β1 knockout mice that died at an early age with a multiorgan inflammatory syndrome (27). The absence of TGF-β1 was associated with increased production of inflammatory cytokines such as IFN-γ and TNF-α, and an increased number of activated immune cells in peripheral lymphoid organs.
Two recent studies (9, 28) demonstrate that TGF-β (β1 or β2) can modulate macrophage activity in a way that favors Th2 over Th1 differentiation. Our results are consistent with these observations, since with TGF-β1 therapy we see suppression of the type 1 proinflammatory cytokines IL-12 and IFN-γ, with minimal alteration of IL-4 gene expression. Based on published data, the regulatory activity of TGF-β1 may result from a direct action on macrophages or may be a consequence of decreased Th1 reactivity. In fact, this cytokine could be blocking several steps of an immune reaction, including Ag processing/presentation by APCs (29), activation/differentiation of Th1 cells (30), production of inflammatory cytokines and NO (30, 31), and activation of effector cells (CTLs, macrophages, NK cells) (32). TGF-β1 blocks JAK-STAT signaling in T cells by preventing tyrosine phosphorylation and activation of Jak-1 and STAT-5 (33). It inhibits IL-2R expression, and can also induce apoptosis in T cells (33). These mechanisms are not mutually exclusive, and it would be difficult to ascertain their relative importance in vivo in NOD mice. Interestingly, the increased number of normal islets in TGF-β1-treated mice suggests that either fewer islet-reactive T cells are activated or that their migration to islets of Langerhans is impeded. The site of activation of latent TGF-β1 is unclear, as discussed below.
Previously, others found that i.m. injection of TGF-β1 expression vectors was therapeutically effective in models of inflammatory bowel disease (34) and systemic lupus erythematosus (35). These investigators administered plasmid vectors with a Rous sarcoma virus enhancer/promoter (compared with a CMV enhancer/promoter in our study), and either did not report an increase in plasma TGF-β1 (34), or observed a smaller increase than in our experiments (35). It appears that we achieved a higher level of TGF-β1 production, possibly because the CMV enhancer/promoter of our vector generates a higher transcriptional activity (5, 8).
Adoptive transfer of TGF-β1-producing islet-reactive CD4+ T cells prevents diabetes in NOD mice (36). However, transgenic NOD mice producing active TGF-β1 in their islets (rat insulin promoter) developed insulitis and marked fibrosis in the pancreas (37). Immunoprotection may depend on the TGF-β1 levels achieved locally or systemically, and/or the state of TGF-β1 activation.
The ability of TGF-β1 to induce fibrosis and extracellular matrix formation (38) is a potential concern. In mice treated with pCMV-TGF-β1, we did not observe fibrosis or inflammation in muscles, kidneys, lungs, liver, heart, or pancreas. We speculate that the latent TGF-β1 is not activated at the site of plasmid administration, since active TGF-β1 was not detectable in plasma. Others have shown (26) that administration of latent TGF-β1 is probably an advantageous feature, since it may become biologically active only at distant inflammatory sites, through the action of macrophages, low pH, or other factors (26).
As mentioned above, IFN-γ may contribute to islet cell injury, since it has direct toxic effects on islet cells, and may also act by activating macrophages and stimulating NO production. In accordance with this view, disease is delayed in IFN-γ gene knockout NOD mice (39), and prevented by anti-IFN-γ mAb treatment (40). To confirm the detrimental effects of IFN-γ, as well as the efficacy of our gene delivery method, we administered an IFN-γ-expressing vector (pCMV-IFN-γ) to CYP-treated NOD mice. The serum levels of IFN-γ were increased in pCMV-IFN-γ-treated mice (up to 200 pg/ml), and this was sufficient to induce an earlier onset of diabetes. Acceleration of this disease by administration of IFN-γ had not been reported before.
To our knowledge, we are the first to demonstrate that IDDM can be prevented by cytokine i.m. somatic gene therapy, and to demonstrate that TGF-β1 is therapeutically effective in this disease. Clinically, the use of cytokines has been limited by their short t1/2 and the necessity to administer relatively large quantities (often in boluses) of recombinant proteins, with considerable associated toxicity (41). Somatic cytokine gene therapy has the potential of circumventing these problems by minimizing the need for frequent protein injections, producing more constant blood levels, reducing side effects, and increasing therapeutic efficacy. Furthermore, direct DNA injection in skeletal muscle appears to be safe and is technically simple. Unlike some viral vectors, the plasmid vectors are maintained episomally (5), minimizing the risk of genomic insertional mutagenesis. Moreover, the use of plasmid vectors eliminates the possibility of immune responses to viral Ags.
Various nonviral means of introducing DNA into cells have been developed. Although cationic liposomes usually enhance DNA uptake by cells, i.v. injection of DNA/liposome complexes has met with limited success. In that case, the cells of several organs are transfected (particularly endothelial cells) (42, 43), but the persistence of gene expression is usually much less than after i.m. delivery (43; and our unpublished observations). The potential toxicity of liposomes must also be considered (43). Thus, the muscle cell is an excellent target for gene therapy, and in these cells injection of naked plasmid DNA has proven to be an effective method.
We thank Dr. François Bellavance (Department of Epidemiology and Biostatistics, McGill University, Montreal, Canada) for his assistance in the statistical analysis of our data.
This study was funded by the Juvenile Diabetes Foundation International.
Abbreviations used in this paper: NOD, nonobese diabetic; IDDM, insulin-dependent diabetes mellitus; CYP, cyclophosphamide; DTH, delayed-type hypersensitivity; mTGF, mouse transforming growth factor; NO, nitric oxide; RF, rectus femoris; TA, tibialis anterior.