Research ArticleType 1 Diabetes

Insulin B chain 9–23 gene transfer to hepatocytes protects from type 1 diabetes by inducing Ag-specific FoxP3+ Tregs

See allHide authors and affiliations

Science Translational Medicine  27 May 2015:
Vol. 7, Issue 289, pp. 289ra81
DOI: 10.1126/scitranslmed.aaa3032


Antigen (Ag)–specific tolerance in type 1 diabetes (T1D) in human has not been achieved yet. Targeting lentiviral vector (LV)–mediated gene expression to hepatocytes induces active tolerance toward the encoded Ag. The insulin B chain 9–23 (InsB9–23) is an immunodominant T cell epitope in nonobese diabetic (NOD) mice. To determine whether auto-Ag gene transfer to hepatocytes induces tolerance and control of T1D, NOD mice were treated with integrase-competent LVs (ICLVs) that selectively target the expression of InsB9–23 to hepatocytes. ICLV treatment induced InsB9–23–specific effector T cells but also FoxP3+ regulatory T cells (Tregs), which halted islet immune cell infiltration, and protected from T1D. Moreover, ICLV treatment combined with a single suboptimal dose of anti-CD3 monoclonal antibody (mAb) is effective in T1D reversal. Splenocytes from LV.InsB9–23–treated mice, but not from LV.OVA (ovalbumin)–treated control mice, stopped diabetes development, demonstrating that protection is Ag-specific. Depletion of CD4+CD25+FoxP3+ T cells led to diabetes progression, indicating that Ag-specific FoxP3+ Tregs mediate protection. Integrase-defective LVs (IDLVs).InsB9–23, which alleviate the concerns for insertional mutagenesis and support transient transgene expression in hepatocytes, were also efficient in protecting from T1D. These data demonstrate that hepatocyte-targeted auto-Ag gene expression prevents and resolves T1D and that stable integration of the transgene is not required for this protection. Gene transfer to hepatocytes can be used to induce Ag-specific tolerance in autoimmune diseases.


Type 1 diabetes (T1D) is an autoimmune disease resulting in the complete destruction of insulin-producing pancreatic β cells (1, 2). In human T1D and in the nonobese diabetic (NOD) mouse, the spontaneous murine model of T1D, autoreactive CD4+ and CD8+ effector T cells have been shown to target islet-associated antigens (Ags), including glutamic acid decarboxylase (GAD), islet-specific glucose 6-phosphotase catalytic subunit–related protein (IGRP), chromogranin A, zinc transporter 8, and insulin (36). Effector T cells express high levels of interferon-γ (IFN-γ), perforin, and granzyme and kill target β cells (7, 8). In T1D, effector responses are predominant over tolerogenic responses, partly due to defects in peripheral regulatory T cells (Tregs) of the host. Tregs keep effector T cells and inflammatory responses in check via cytokine- and cell-to-cell contact–mediated mechanisms (9). Therefore, a dysfunction in Tregs allows effector T cells to prevail and autoaggression to ensue. Reduced frequency and function of CD4+ FoxP3+ Tregs with expansion of autoreactive effector T cells have been reported in mouse and humans with T1D (10, 11). An attractive approach to restore peripheral tolerance to islet Ags and prevent T1D onset is to induce and/or expand islet-specific Tregs that can control autoreactive effector T cells (12, 13).

In the past decade, encouraging results showed that Ag-based immunotherapy could be used to restore tolerance in T1D. Several islet Ags including insulin, GAD, heat shock protein 60, and others have been administered to prevent disease onset (14, 15). Despite promising preclinical data showing disease prevention, the Ag-based therapies have largely failed to show efficacy in the clinic (16, 17). A definitive explanation for these failures in humans is lacking, but it appears that the immune response elicited by administration of the auto-Ag is not protective. The type and robustness of the immune response induced by auto-Ags depend on the site and context of Ag presentation to the T cells. The tolerogenic versus inflammatory properties of certain organs make Ag delivery strategies a critical factor in designing efficacious immunotherapies for the treatment of autoimmune diseases.

One promising approach to induce Ag-specific Tregs is the direct delivery of the Ag to the liver. The tolerogenic properties of the liver have been extensively demonstrated (1820). Constitutive production of immunosuppressive cytokines such as interleukin-10 and transforming growth factor–β has been reported (21). In addition, Ag presentation by liver resident cells results in tolerogenic T cell responses (2225). To explore the possibility of exploiting the liver to induce Tregs specific for the immunodominant epitope of insulin [insulin B chain 9–23 (InsB9–23)], which are capable of controlling T1D development, we designed a lentiviral vector (LV), which selectively targets expression of the transgene to the hepatocytes. This LV includes two regulatory elements in the expression cassette: hepatocyte-specific promoter [enhanced transthyretin (ET)] for positive transcriptional regulation, and micro-RNA142 target sequences (142T) for negative posttranscriptional regulation in hematopoietic cell lineage (LV.ET.142T) (26, 27). We previously showed that expression of transgenes encoded by this LV design in hepatocytes leads to the induction of an active state of immune tolerance mediated by FoxP3+ Tregs specific for a neo-Ag (2830).

Here, we demonstrate that LV.ET.142T-mediated gene transfer suppresses ongoing T1D autoimmune responses by expressing an early auto-Ag–derived epitope (InsB9–23) in hepatocytes. Reversal of overt T1D can be achieved by gene transfer combined with a single suboptimal dose of anti-CD3 monoclonal antibody (mAb) (31, 32). The treatment resulted in enrichment of FoxP3+ Tregs in the liver, pancreatic lymph nodes (PLNs), and pancreatic islets. These InsB9–23–specific FoxP3+ Tregs suppress T cell–mediated diabetogenic responses against multiple auto-Ags, arresting T1D progression.


Hepatocyte-targeted InsB9–23 expression blocks T1D progression

We injected systemically a series of LV constructs encoding for InsB9–23 in 10-week-old NOD mice. Glycemia was monitored to determine the efficacy in controlling diabetes progression. Integrase-competent LVs (ICLVs) carrying the ubiquitously active promoter [phosphoglycerate kinase 1 (PGK)] to drive InsB9–23 expression (ICLV.PGK.InsB9–23) in all transduced cell types failed to block T1D development (Fig. 1A). An ICLV.PGK vector, which incorporates micro-RNA142 target sequences (142T) (ICLV.PGK.InsB9–23.142T) to negatively regulate transgene expression in hematopoietic lineage cells, was also unable to stop β cell destruction (Fig. 1B). To selectively target InsB9–23 expression to hepatocytes, we used an LV design in which we combined the ET hepatocyte-specific promoter and 142T regulatory elements (ICLV.ET.InsB9–23.142T). Ninety percent of NOD mice injected with ICLV.ET.InsB9–23.142T were protected from disease development and remained normoglycemic up to 43 weeks of age (mean blood glucose levels: 135 ± 14 mg/dl) (Fig. 1C). NOD mice treated with control vectors encoding for an unrelated Ag, such as ovalbumin (OVA) (ICLV.PGK.OVA.142T and ICLV.ET.OVA.142T), failed to protect NOD mice from T1D, although the hepatocyte-targeted OVA vector induced some delay in T1D development (Fig. 1, B and C). Overall, these results indicate that the expression of InsB9–23 must be stringently targeted to the liver parenchyma for maintenance of normoglycemia and T1D protection.

Fig. 1. Liver-directed InsB9–23 expression protects from T1D.

(A to C) Ten-week-old NOD females at late prediabetic stage (blood glucose 100 ± 23 mg/dl, n = 36) were treated with a single dose of ICLV by systemic injection of (A) ICLV.PGK.InsB9–23 (n = 8), (B) ICLV.PGK.InsB9–23.142T (n = 8), ICLV.PGK.OVA.142T (n = 5), and (C) ICLV.ET.InsB9–23.142T (n = 10; ****P < 0.0001, Kaplan-Meier log-rank test) and ICLV.ET.OVA.142T (n = 5) or left untreated as control [n = 7, 10, and 14 for (A), (B), and (C), respectively]. Blood glucose levels were measured to monitor T1D progression in NOD mice, which were considered diabetic when the glucose levels were above 250 mg/dl. (D) Immunohistochemical analysis of pancreatic islets was performed on sections obtained from normoglycemic NOD untreated control mice (n = 4, 10-week-old mice, blood glucose 100 ± 23 mg/dl, n = 100 islets; n = 6, 16-week-old mice, blood glucose 203 ± 15 mg/dl, n = 140 islets) and from NOD mice injected at 10 weeks of age with ICLV.ET.OVA.142T (n = 4 at 16 weeks of age, blood glucose 205 ± 32 mg/dl, n = 90 islets) or with ICLV.ET.InsB9–23.142T (n = 6 at 16 weeks of age, blood glucose 119 ± 15 mg/dl, n = 140 islets; n = 6 at 43 weeks of age, blood glucose 135 ± 14 mg/dl, n = 236 islets; ****P < 0.0001, χ2 test versus ICLV.ET.OVA.142T-treated at 16 weeks or versus untreated controls) to determine the level of lymphocytic infiltration. Data are shown as percentage of islet with a given level of infiltration (no infiltration; peri-infiltrated, infiltrated less than 50%, and infiltrated more than 50% in the islet area). (E) Representative images of anti-CD3 and anti-insulin immunostaining are reported (scale bar, 50 μm).

Mononuclear cells infiltrating pancreatic islets are responsible for β cell death, and the extent of islet infiltration (insulitis) correlates with the progression of T1D in NOD mice (33). To assess the degree of insulitis, we performed histological analysis of pancreatic tissue isolated from control NOD mice (either untreated or treated with ICLV.ET.OVA.142T) and from ICLV.ET.InsB9–23.142T–treated NOD mice. Results indicate that in 10-week-old NOD mice (at the time of LV administration), ~25% of the islets were heavily infiltrated. The progression of the disease was evident in 16-week-old, untreated NOD mice, which displayed ~80% of islets severely infiltrated. A comparable infiltration pattern occurred in 16-week-old NOD mice 6 weeks after ICLV.ET.OVA.142T treatment, confirming that OVA expression in hepatocytes does not halt β cell death. On the contrary, at 16 weeks of age, the ICLV.ET.InsB9–23.142T–treated mice displayed an insulitis profile similar to that of 10-week-old NOD mice: Only ~20% of islets were heavily infiltrated, whereas the remaining islets showed minimal CD3+ T cell infiltration and normal insulin production. Comparable data were obtained in ICLV.InsB9–23.142T–treated NOD mice at 43 weeks of age (Fig. 1, D and E). These results indicate that InsB9–23 ICLV–mediated expression in hepatocytes blocks infiltration by diabetogenic effector T cells and allows the maintenance of insulin production.

Hepatocyte-targeted InsB9–23 expression results in long-term transgene expression despite the presence of Ag-specific effector CD8+ T cells

To investigate the effects of ICLV.ET.InsB9–23.142T administration on adaptive immunity in mice with an ongoing autoimmune response, we tested the presence of InsB9–23–specific cytotoxic CD8+ T lymphocytes (CTLs). The frequency of CTLs responsive to InsB was quantified in the spleen and in the liver of ICLV.PGK.InsB9–23– and ICLV.PGK.InsB9–23.142T–treated NOD mice at 16 weeks (fig. S1) and ICLV.ET.InsB9–23.142T–treated NOD mice at 43 weeks of age (Fig. 2, A and B). As control, the InsB-specific response was tested in overt diabetic mice (glucose >300 mg/dl) and in 10-week-old normoglycemic untreated NOD mice. In parallel, OVA-specific responses were investigated in ICLV.ET.OVA.142T- or ICLV.ET.InsB9–23.142T–treated mice, and in overt diabetic and normoglycemic untreated NOD mice (Fig. 2, A and B). Data show that InsB9–23 gene transfer induced a significant increase in CD8+ T cells in the liver (Fig. 2A). These cells release IFN-γ upon in vitro stimulation with a transgene-expressing cell line (Fig. 2, A and B) and kill InsB15–23–pulsed target cells as much as CTLs derived from diabetic mice (fig. S2), demonstrating their competence in exerting effector functions. The insulin-specific response increased also in the spleen of ICLV.ET.InsB9–23.142T–treated NOD mice, and it was significantly higher in comparison to that observed in mice with overt disease or in untreated mice. On the other hand, the response to OVA in ICLV.ET.InsB9–23.142T–treated NOD mice was comparable to that detected in T1D or untreated mice. In ICLV.ET.OVA.142T-treated mice, a strong CTL response to OVA was observed in the liver and in the spleen, whereas the response to InsB9–23 was very low.

Fig. 2. Tolerance to transgene is achieved with liver-directed expression despite induction of transgene-specific CTLs.

(A and B) IFN-γ–secreting, InsB/OVA-specific CD8+ T cells in the liver (A) and in the spleen (B) were quantified by ELISPOT (enzyme-linked immunospot) assay in NOD mice treated with ICLV.ET.InsB9–23.142T [n = 4 liver, n = 3 spleen; ***P < 0.001, analysis of variance (ANOVA)] and ICLV.ET.OVA.142T (n = 2 liver and spleen; ***P < 0.001, ANOVA), or in untreated 10-week-old (n = 6 liver, n = 7 spleen) and fully diabetic (n = 3 liver and spleen, blood glucose >300 mg/dl) NOD mice as control. Data are presented as means ± SEM of InsB/OVA-specific CD8+ T cells out of 106 total CD8+ T cells. (C) VCN per diploid genome in the liver of NOD mice measured by quantitative polymerase chain reaction (**P < 0.01, ANOVA). VCN mean ± SEM is shown.

Despite the induction of CTLs directed toward the InsB9–23 transgene, high LV genome copies [vector copy number (VCN)] were observed in the liver of ICLV.ET. InsB9–23.142T–treated NOD mice, suggesting that LV-transduced cells expressing InsB9–23 are protected from the immune-mediated clearance. Similarly, in ICLV.ET.OVA.142T-treated mice, high VCN was observed in the liver despite a strong response to OVA by CTLs. On the contrary, the injection of an equal dose of ICLV.PGK.InsB9–23 or ICLV.PGK.InsB9–23.142T constructs resulted in a very low VCN in the liver of treated NOD mice, although integrated LV genomes were higher in the liver of ICLV.PGK.142T-treated NOD mice (Fig. 2C).

Together, these data show that gene therapy with ICLV.ET.142T containing either InsB or OVA induces and/or boosts CTL responses toward the transgene. However, transgene expression restricted to hepatocytes prevents clearance of transduced cells.

InsB9–23–specific Tregs are induced after treatment with ICLV.ET.InsB9–23.142T

To investigate if active tolerance plays a role in protecting β cells from clearance mediated by CTLs, we evaluated the frequency of FoxP3+ Tregs in the spleen, liver, and PLNs of mice 6 weeks after ICLV administration (Fig. 3A). The percentage of FoxP3+ Tregs in the PLN of ICLV.ET.InsB9–23.142T–treated mice was significantly higher as compared to that of ICLV.ET.OVA.142T-treated or untreated NOD mice (Fig. 3A). Higher frequency of FoxP3+ Tregs was also detected in the spleen of ICLV.ET.InsB9–23.142T–treated mice in comparison to untreated NOD mice (Fig. 3A). On the other hand, the frequency of FoxP3+ Tregs in the liver of both ICLV.ET.InsB9–23.142T– and ICLV.ET.OVA.142T–treated NOD mice was significantly increased compared to what we observed in untreated control mice (Fig. 3A). These data indicate that ICLV.ET.142T administration expands FoxP3+ Treg population, suggesting induction of tolerance to the encoded Ag.

Fig. 3. FoxP3-expressing Tregs are up-regulated in mice receiving liver-directed LV.

(A) The frequency of FoxP3+ Tregs was determined by FACS (fluorescence-activated cell sorting) in the spleen, liver, and PLN of NOD mice at 6 weeks (representative FACS plots for FoxP3 staining in the PLN are shown) after injection with ICLV.ET.InsB9–23.142T (n = 8; **P < 0.01, ***P < 0.001, ****P < 0.0001, ANOVA) and ICLV.ET.OVA.142T (n = 5; *P < 0.05, ****P < 0.0001, ANOVA) or of untreated 10-week-old NOD mice (n = 5). (B) Immunohistochemical analysis of FoxP3+ cells in pancreatic islets was performed on sections obtained from 10-week-old normoglycemic NOD untreated mice (n = 4, n = 26 islets) and from 16-week-old NOD mice injected at 10 weeks of age with ICLV.ET.OVA.142T or with ICLV.ET.InsB9–23.142T (n = 4, n = 21 islets; n = 6, n = 22 islets; ***P < 0.001, ANOVA) or at 43 weeks of age (n = 2, n = 16 islets; n = 6, n = 26 islets; ***P < 0.001, ANOVA). The number of FoxP3+ cells was determined and data are reported as mean of FoxP3+ cell sorting for grade of infiltration ± SEM. (C) Representative images are reported for 16-week-old untreated and 43-week-old ICLV.ET.InsB9–23.142T–treated NOD mice (scale bar, 50 μm). Splenocytes, liver-infiltrating lymphocytes, and PLN cells were isolated from ICLV.ET.InsB9–23.142T–treated (n = 2, 16-week-old) NOD mice and pooled to magnetically sort CD4+CD25+ T cells, which are highly enriched in FoxP3+ Tregs. CD4+CD25 responder T cells were isolated and pooled from diabetic NOD mice (TT1D) (n = 2). (D and E) The capacity of enriched FoxP3+ Tregs to suppress T cell responses was tested after polyclonal stimulation (anti-CD3 mAb plate bound, 10 μg/ml) (D) or after Ag-specific stimuli (InsB9–23 or IGRP195–214 peptides, 10 μg/ml) (E). Proliferation of TT1D cells was quantified by [3H]thymidine incorporation. Percentage of suppression is shown. Mean CPM × 10−3 ± SEM for each experimental condition is reported.

The preferential accumulation of FoxP3+ Tregs in the spleen, liver, and PLN of ICLV.ET.InsB9–23.142T–treated mice correlates with the gene expression data showing an increased expression of Treg-associated markers and immunoregulatory cytokines in CD4+ splenocytes, liver-infiltrating lymphocytes, and PLN cells (foxp3, ctla4, il10, and tgfβ; fig. S3). This can be due to the presence of the relevant auto-Ag in PLN and spleen, in addition to its expression in the liver, whereas expression of OVA is confined to the liver.

To determine whether FoxP3+ Tregs were also increased in the pancreatic islets, FoxP3+-expressing cells were counted in the infiltrated areas of pancreatic islets obtained from mice treated with either ICLV.ET.InsB9–23.142T or ICLV.ET.OVA.142T at 6 and 33 weeks after LV treatment and untreated controls. Immunohistochemical analysis identified FoxP3+ Tregs as a significantly increased population in the islets with <50 and >50% infiltration in ICLV.ET.InsB9–23.142T–treated mice at both time points (Fig. 3, B and C).

To evaluate whether ICLV treatment selectively induces FoxP3+ Tregs, we measured the expression of T helper 1 (TH1)–, TH2-, and TH17-specific cytokines and transcription factors (ifn-γ, tbet, il4, gata 3, il17, rorc) in CD4+ splenic T cells and PLN cells. Results showed that none of these T cell subsets were preferentially expanded by ICLV.ET.InsB9–23.142T or ICLV.ET.OVA.142T treatment (fig. S4).

The suppressive activity of CD4+CD25+FoxP3+ Tregs isolated from 16-week-old ICLV.ET.InsB9–23.142T–treated NOD mice was tested in vitro in coculture with CD4+ T effector cells isolated from diabetic mice. Tregs suppressed proliferation of effector T cells more efficiently after activation with InsB9–23 peptide than with IGRP195–214 peptide (Fig. 3, D and E).

Overall, the distribution of Tregs within the liver, PLN, and pancreas after ICLV.ET.142T treatment, together with the ability of the liver to generate transgene-specific Tregs (28), indicates that InsB9–23–specific Tregs are induced in the liver and accumulate in the peripheral anatomical sites where the cognate Ag is presented to T cells.

Active protection from T1D induced by ICLV.ET.InsB9–23.142T treatment is Treg-dependent

To determine whether protection from T1D is due to an active mechanism of Ag-specific immune regulation mediated by FoxP3+ Tregs, splenocytes isolated from 16-week-old ICLV.ET.InsB9–23.142T– or ICLV.ET.OVA.142T–treated NOD mice were cotransferred with splenocytes isolated from diabetic NOD mice in NOD.scid mice. Significant protection from T1D was observed in the group cotransferred with splenocytes from ICLV.ET.InsB9–23.142T–treated mice (Fig. 4A). On the other hand, T1D onset occurred at a similar rate in NOD.scid mice repopulated with splenocytes from diabetic mice alone or cotransferred with splenocytes from ICLV.ET.OVA.142T-treated NOD mice. The protection mediated by splenocytes isolated from ICLV.ET.InsB9–23.142T–treated NOD mice was completely abrogated when CD4+CD25+ cells were depleted from the splenocytes before the cotransfer (Fig. 4B). The depletion procedure removed 75% of CD4+CD25+FoxP3+ splenocytes (fig. S5).

Fig. 4. T1D protection in InsB9–23 mice is Treg-dependent.

(A) Splenocytes isolated from diabetic NOD mice were cotransferred into NOD.scid mice at 1:1 ratio with splenocytes isolated from ICLV.ET.InsB9–23.142T–treated (n = 6, 16-week-old; *P < 0.05, **P < 0.01, Kaplan-Meier log-rank test) and ICLV.ET.OVA.142T-treated (n = 5, 16-week-old) NOD mice or transferred alone. (B) To evaluate the role of FoxP3+ Tregs, which are mainly included in the CD4+CD25+ T lymphocytes, splenocytes isolated from diabetic NOD mice were cotransferred into NOD.scid mice at 1:1 ratio with CD4+CD25+-depleted splenocytes isolated from ICLV.ET.InsB9–23.142T–treated (n = 6, 16-week-old) or ICLV.ET.OVA.142T-treated (n = 5, 16-week-old) NOD mice. Blood glucose levels were measured to monitor T1D transfer in NOD.scid mice. Diabetes incidence is shown.

To define whether Ag-specific Tregs induced by ICLV.ET.InsB9–23.142T treatment also exert bystander suppression toward other β cell–specific Ags, we evaluated the IGRP206–214–specific CD8+ T cell response in diabetic and age-matched ICLV.ET.InsB9–23.142T–treated NOD mice. Results indicate that ICLV.ET.InsB9–23.142T treatment also reduced IGRP206–214–specific CD8+ T cell response (fig. S6). Overall, these results demonstrate that the suppression of diabetogenic responses by ICLV.ET.InsB9–23.142T treatment is mediated by Ag-specific CD25+FoxP3+ Tregs, which also exert bystander suppression.

Integrase-defective LV InsB9–23 gene transfer protects from T1D

The use of LVs as a vaccine-based platform to generate immune tolerance may raise concerns due to the risks of insertional mutagenesis, although extensive preclinical and clinical studies support the safety of LV-mediated gene transfer (3437). Moreover, long-term transgene expression may not be required to maintain immune tolerance, in particular to auto-Ags, which are present in the organism. We previously demonstrated that the intravenous administration of an integrase-defective LVs (IDLVs) encoding for green fluorescent protein (GFP), IDLV.ET.GFP.142T, led to transient expression of GFP in hepatocytes while retaining the ability to induce a robust state of Ag-specific tolerance (29).

To determine whether liver-directed, IDLV-mediated InsB9–23 gene transfer arrests T1D, 10-week-old NOD mice were treated with IDLV.ET.InsB9–23.142T. Blood glucose levels were monitored for 15 weeks after IDLV.ET.InsB9–23.142T treatment, revealing that transient expression of InsB9–23 by hepatocytes suppresses T1D development in 80% of treated NOD mice (mean blood glucose: 127 ± 49 mg/dl) (Fig. 5A). Histological analysis of the pancreata revealed that IDLV.ET.InsB9–23.142T treatment inhibited the progressive infiltration of the pancreatic islets (Fig. 5B). On the contrary, in IDLV.ET.OVA.142T-treated mice, diabetes occurred as in untreated control mice. VCN analysis confirmed absent to very low IDLV genomes in the liver of treated mice at the end of experiment (fig. S7C). Similar to what was observed after ICLV treatment, transgene-specific effector cells were detected in the spleen and liver after IDLV.ET.142T gene transfer (fig. S7, A and B), and the frequency of FoxP3+Tregs was increased in the PLN and in pancreatic islets only in mice treated with IDLV.ET.InsB9–23.142T (Fig. 6A and figs. S8 and S9). Cotransfer of splenocytes isolated from IDLV.ET.InsB9–23.142T–treated mice with splenocytes from diabetic NOD mice demonstrated that IDLV.ET.InsB9–23.142T treatment induces Ag-specific active immunoregulation of diabetogenic T cells (Fig. 6B). Moreover, depletion of CD4+CD25+ splenocytes before cotransfer demonstrated that Ag-specific active immunoregulation was dependent on the activity of CD4+CD25+FoxP3+ Tregs (Fig. 6C), as shown for ICLV treatment.

Fig. 5. Integrase-defective InsB9–23 LV treatment protects from T1D.

(A) Ten-week-old NOD females at late prediabetic stage were treated with single dose of IDLVs by systemic injection of IDLV.ET.InsB9–23.142T (n = 6; *P < 0.05, Kaplan-Meier log-rank test) and IDLV.ET.OVA.142T (n = 4) or left untreated as control (n = 15). Blood glucose levels were measured to monitor T1D progression. Diabetes incidence is shown. (B) Immunohistochemical analysis of pancreatic islets was performed on sections obtained from normoglycemic NOD untreated control mice (n = 4, 10-week-old mice, n = 63 islets; n = 4 16-week-old mice, n = 62 islets) and from NOD mice injected at 10 weeks of age with IDLV.ET.OVA.142T (n = 4, 16-week-old mice, blood glucose 202 ± 18 mg/dl, n = 49 islets) or with IDLV.ET.InsB9–23.142T (n = 5, 16 weeks of age, glucose 144 ± 19 mg/dl, n = 85 islets; n = 4, 35 weeks of age, blood glucose 138 ± 23 mg/dl, n = 119 islets) to determine the level of lymphocytic infiltration. Data are shown as percentage of islets with a given level of infiltration (no infiltration; peri-infiltration, infiltration less than 50%, and infiltration more than 50% of the islet area).

Fig. 6. IDLVs induce Ag-specific Tregs capable of controlling T1D.

(A) The frequency of FoxP3+ Tregs was determined by FACS in spleen, liver, and PLN 15 weeks after treatment of IDLV.ET.InsB9–23.142T (n = 5; **P < 0.01, ***P < 0.001, ANOVA) and IDLV.ET.OVA.142T (n = 3; *P < 0.05, **P < 0.01, ANOVA). Untreated 10-week-old NOD mice (n = 3) were used as control group. Data are presented as means ± SEM. (B) Splenocytes isolated from diabetic NOD mice were cotransferred into NOD.scid mice at 1:1 ratio with splenocytes isolated from IDLV.ET.InsB9–23.142T–treated (n = 4, 16-week-old, **P < 0.01, Kaplan-Meier log-rank test) and IDLV.ET.OVA.142T-treated (n = 4, 16-week-old) mice or transferred alone (n = 4). (C) To evaluate the role of FoxP3+ Tregs, which are mainly included in the CD4+CD25+ T lymphocytes, splenocytes isolated from diabetic NOD mice were cotransferred into NOD.scid mice at 1:1 ratio with CD4+CD25+-depleted splenocytes isolated from IDLV.ET.InsB9–23.142T–treated (n = 6, 16-week-old) or IDLV.ET.OVA.142T–treated (n = 4, 16-week-old) NOD mice. Blood glucose levels were measured to monitor T1D transfer in NOD.scid mice. Diabetes incidence is shown.

Hepatocyte-targeted InsB9–23 expression combined with suboptimal dose of anti-CD3 reverses T1D

The efficacy of hepatocyte-directed InsB9–23 gene transfer in controlling T1D was also tested at later stages of the disease progression. ICLV.ET.InsB9–23.142T treatment administered in NOD mice at the end of the presymptomatic phase when glycemic levels range from 200 to 250 mg/dl protected 27% of the mice (fig. S10A). We also tested ICLV.ET.InsB9–23.142T treatment in diabetic NOD mice with blood glucose levels ranging from 250 to 300 mg/dl; none of the treated mice were cured with reversal to normoglycemic levels (fig. S10B). We next combined InsB9–23 gene transfer with anti-CD3 mAb treatment. Treatment with optimal doses of anti-CD3 mAb alone can reverse T1D in NOD mice (31, 32). Therefore, to identify the suboptimal dose of anti-CD3 mAb with no protective effect, decreasing doses of anti-CD3 mAb were tested in diabetic NOD mice with blood glucose levels ranging from 250 to 300 mg/dl. We observed that a single administration of anti-CD3 mAb at 5 μg per mouse instead of 10 μg per mouse was not effective in protecting from T1D (fig. S11).

On the basis of these results, we treated NOD mice with blood glucose level ranging from 250 to 300 mg/dl with this suboptimal dose of anti-CD3 mAb (1 × 5 μg) together with ICLV.ET.InsB9–23.142T. This treatment achieved 75% T1D reversal (9 of 12 mice tested; Fig. 7). In 4 of the reverted mice, we observed temporary spikes of blood glucose levels, which returned below the 250 mg/dl threshold, indicating that a significant portion of β cell mass was still present and functional.

Fig. 7. ICLV.ET.InsB9–23.142T combined with a suboptimal dose of anti-CD3 reverts T1D.

Diabetic NOD mice with blood glucose levels ranging from 250 to 300 mg/dl (gray area) were treated with ICLV.ET.InsB9–23.142T combined with anti-CD3 mAb (1 × 5 μg) (n = 12). Blood glucose levels of each single mouse are reported starting from the time of treatment (color lines). Mice were considered as reverted when the glucose levels returned below 250 mg/dl. Red dashed lines identify NOD mice with blood glucose levels >600 mg/dl, which is consistent with the absence of insulin-secreting β cells.

These data indicate that treatment with a suboptimal dose of anti-CD3 mAb is sufficient to reestablish a permissive condition for tolerance induction mediated by hepatocyte-targeted gene transfer.

Overall, these results demonstrate that the ectopic expression of auto-Ags in hepatocytes generates Ag-specific CTL but also Tregs, which actively suppress the already ongoing diabetogenic responses in presymptomatic phase and reverse T1D, in combination with a suboptimal dose of anti-CD3 mAb. The comparable efficacy of the ICLV and IDLV platforms demonstrates that the stable persistence of transgene expression is not required to achieve long-lasting Ag-specific immune tolerance.


A growing body of evidence indicates that the autoimmune attack underlying β cell destruction does not lead to clinical symptoms of T1D for many years in most patients. At clinical onset, T1D patients have already lost more than 80% of β cell mass, and therefore, an effective therapy should specifically suppress ongoing diabetogenic responses while preserving the remaining insulin-producing β cells.

Here, we show that LV-based gene transfer targeting auto-Ag expression to hepatocytes protects NOD mice from T1D. Systemic administration of a single dose of ICLV.ET.InsB9–23.142T arrested β cell destruction in NOD at advanced prediabetic stage, “freezing” islet infiltration at the stage observed at the time of treatment, and maintaining insulin independence in 90% of LV-treated NOD mice. Targeting InsB9–23 expression to hepatocytes was required to generate InsB9–23–specific FoxP3+ Tregs, which allowed the maintenance of stable normoglycemia, suppressing diabetogenic T cell responses at the site of cognate Ag presentation.

We previously showed that hepatocyte-directed gene therapy by ICLVs encoding for factor IX (FIX) eradicated preexisting anti-FIX–neutralizing response, and provided therapeutic levels of the missing clotting factor in a mouse model of hemophilia B (FIX-deficient), resolving the disease (30). Here, we developed a hepatocyte-directed LV gene transfer strategy, which exploits naturally occurring hepatic tolerogenic pathways to suppress immune responses to auto-Ags. We demonstrate that hepatocyte-directed ICLV-mediated auto-Ag (InsB9–23) gene transfer efficiently suppresses ongoing autoimmune responses against a variety of self-Ags expressed by pancreatic β cells, arresting T1D development. Gene transfer using IDLVs led to comparable results, indicating that persistent high levels of InsB9–23 expression in hepatocytes are not required for induction of tolerance to self-Ags and long-term maintenance of insulin production. Moreover, hepatocyte-directed gene transfer resulted in a significant reversal of T1D when combined with a suboptimal, per se ineffective, dose of anti-CD3 mAb.

Several gene transfer–based approaches to prevent/treat T1D have been investigated (3844). Administration of AAV (39, 40) or mRNA (41) encoding for cytokines resulted in prevention of T1D. In addition, antisense oligonucleotides have been used ex vivo to down-regulate expression of costimulatory molecules in DC, which control autoimmunity once reinfused in vivo (45). All these approaches are non–Ag-specific and therefore may lead to general immune suppression.

Ag-specific immunoregulatory strategies based on ectopic expression of an auto-Ag–derived fragment in certain tissue (that is, muscle, liver) have been explored (38, 46). Han et al. (38) showed that AAV-mediated GAD65500–585 gene transfer to the muscle of NOD mice resulted in prevention of T1D in 80% of mice treated at 7 weeks of age when the autoimmune disease is at the very early stage. Lüth et al. (46) showed that myelin basic protein expression in the liver protects from autoimmune neuroinflammation in a mouse model of multiple sclerosis.

Ag-specific prevention of diabetogenic responses has been shown using plasmid DNA encoding for auto-Ags (4143). Recently, pro-insulin gene transfer by plasmid DNA administration has been evaluated in patients, showing safety and an increase in residual C-peptide associated with a decline of pro-insulin–reactive CD8+ T cells (44). However, in this study, an active state of tolerance, which could ensure stable protection from β cell destruction, was not demonstrated.

Multiple routes of administration of insulin or its InsB9–23 immunodominant epitope have been used to prevent T1D (4749). However, insulin-based gene delivery in NOD mice has shown controversial results with variable levels of protection (41, 42, 50, 51). A common feature among the gene delivery systems showing at least partial protection from T1D is the ectopic expression of auto-Ag, leading to the release of immunoregulatory cytokines by T cells and/or the induction of Tregs specific for the encoded Ag (52).

Expression of an auto-Ag in cells of NOD mice is a particular concern because activation of InsB9–23–specific effector T cells could exacerbate β cell autoimmunity, in addition to promoting clearance of transgene-expressing cells. Treatment of mice with LV carrying only the ubiquitous PGK promoter, which permits the expression of the InsB9–23 transgene in all cell types including Ag-presenting cells (APCs), led to induction of effector CTLs, clearance of transduced cells, and development of T1D. Although the ICLV.PGK.InsB9–23 induced a potent immune response, it did not accelerate disease, suggesting that the boost of effector T cell response to insulin induced by this LV construct did not have an additive effect on the already ongoing robust autoimmune responses. The insertion of microRNA-142 target sequences into the vector, to abrogate transgene expression in professional APCs and other hematopoietic lineage cells (27, 28), was not sufficient to confer immunomodulatory properties to the LV construct, which failed to prevent T1D, and it induced transgene-specific CTLs, albeit at a reduced level compared to ICLV.PGK.InsB9–23. The immune response observed after ICLV.PGK.InsB9–23.142T administration is likely the result of the PGK promoter, which expresses the transgene at higher levels than ET promoter in APCs, thus giving rise to some residual transgene expression even if subjected to regulation by microRNA-142 and triggering an immune response to the encoded Ag (26). Transgene expression in APCs, even at very low levels, can inhibit tolerance induction in NOD mice, in which APCs have enhanced immune functions due to hyperactivation of the nuclear factor κB pathway (53).

Replacement of PGK with the hepatocyte-specific ET promoter selectively targeted transgene expression to hepatocytes and diminished the immunogenicity of the LV construct (54, 55). In LV.ET.142T-treated mice, the expression of the transgene is strictly limited to liver parenchyma, and therefore, the priming of effector CTLs is likely mediated directly by the transduced hepatocytes, which can provide only partial costimulation to major histocompatibility complex class I Ag presentation, resulting in suboptimal priming of T cells (56). Indeed, a significant reduction in the absolute number of Ag-specific effector T cells was observed after ICLV.ET.InsB9–23.142T treatment, compared to those observed in ICLV.PGK.InsB9–23 and ICLV.PGK.InsB9–23.142T–treated mice.

FoxP3+ Tregs play a key role in protection from T1D development, because they suppress pathogenic effector T cells (9, 57). Indeed, the induction or adoptive transfer of Tregs has been shown efficacious in the control of ongoing β cell autoimmunity in NOD mice (5860). Whether Ag specificity is required for Tregs to prevent T1D remains an open debate. Here, we show that expression of an early auto-Ag–derived epitope (InsB9–23) in hepatocytes induces Ag-specific Tregs, which mediate robust long-lasting tolerance. Moreover, our study does not suggest any regulatory role of InsB-specific CD8+ T cells, as shown by Santamaria’s group (61). LV.ET.InsB9–23.142T gene transfer led to a preferential expansion of Tregs in the liver, and accumulation in the PLN and endocrine pancreas, where the relevant Ag is present. On the contrary, LV.ET.OVA.142T-treated mice showed no significant variation in the percentage of Tregs in the pancreas and in the PLN, whereas Tregs were expanded in the liver. These data indicate that InsB9–23–specific Tregs accumulated at the site of cognate Ag presentation, whereas OVA-specific Tregs were not attracted to the PLN and pancreas of NOD mice because of lack of the cognate Ag at these sites.

Once Tregs, induced in the liver by LV.ET.InsB9–23.142T therapy, are present at the site(s) where the auto-Ag is presented, they can control autoreactive responses mediated by the same as well as other auto-Ags. Autoimmunity is well established in NOD mice at 10 weeks of age (62), and epitope spreading is already under way at this time (63). Therefore, we hypothesize that the InsB9–23–specific Tregs, which home to the site of Ag expression, regulate autoimmune responses directed toward the same Ag but also different auto-Ags by bystander non–Ag-specific suppression or by inducing in situ and de novo generation of Tregs.

The mechanism underlying the induction of Tregs in the liver is still under investigation. Effector T cells themselves may play a crucial role in the expansion of Tregs. Grinberg-Bleyer et al. (64) showed that effector T cells can boost the number of Tregs, enhancing their regulatory activity in the PLN and pancreas. The induction of transgene-specific effector T cells in the liver after LV.ET.142T administration may therefore determine the frequency of Tregs. It is possible that this regulatory feedback mechanism aimed at protecting InsB9–23–expressing hepatocytes from a massive immune-mediated clearance may also lead to the protection of β cells from diabetogenic responses. On the other hand, the expression of the transgene product in hepatocytes may lead to suboptimal T cell priming, which favors tolerogenic presentation to CD4+ T cells and Treg induction.

The clinical application of ICLVs for hepatocyte-targeted Ag expression to induce immune modulation may be limited by the concerns associated with integration of the vector into the hosts’ cell genome. IDLVs are emerging as an attractive platform for transgene expression [reviewed in (65)]. IDLVs offer the possibility to express transgenes for a window of time in hepatocytes while sharply reducing the integration into the genome. The nonintegrating feature of the IDLV platform provides important safety advantages due to the negligible genotoxic risk and the reversibility of transgene expression. Although the levels of expression achieved with IDLVs are lower compared to those generated by ICLVs, we showed that tolerance to FIX persisted even when the transgene expression became barely detectable (29). Here, the IDLV platform was efficient as its integrating counterpart in controlling T1D, generating Ag-specific tolerance through the induction of InsB9–23-specific Tregs, and driving migration of Tregs to distant anatomical locations to control immune responses.

In conclusion, our study demonstrated that a single LV treatment (LV.ET.InsB9–23.142T) in NOD mice in advanced prediabetes state suppresses immune responses to β cell auto-Ags, saves islet β cell mass, and induces long-lasting immune tolerance. Moreover, the IDLV-based therapy showed the same tolerogenic properties as the ICLV platform, giving the advantage of avoiding any permanent genetic changes of target cells. Large-scale Good Manufacturing Practices (GMP) manufacturing of LVs is now well established, and their clinical testing for the gene therapy of other diseases has been proven safe and successful until now, thus paving the way to broader applications (36, 37).

Multiple genetic and immunological factors, which have been identified as predictive biomarkers for T1D (6670), allow the selection of high-risk patients, in whom intervention when autologous β cell mass is still sufficient can result in maintenance of insulin independence. For these subjects, hepatocyte-targeted gene transfer may represent an effective personalized therapeutic approach, although not without ethical considerations. The efficacy of LV gene transfer, combined with a suboptimal dose of anti-CD3 mAb, to revert T1D and induce Ag-specific tolerance significantly increases the feasibility of its clinical application.


Study design

To study the efficacy of hepatocyte-targeted InsB9–23 expression to arrest and cure T1D, NOD mice were injected intravenously with ICLV or IDLV encoding for InsB9–23. Mice were treated either at 10 weeks of age to arrest T1D progression or when blood glucose levels ranged from 250 to 300 mg/dl to revert hyperglycemia and cure the disease. Experimental groups were dimensioned to allow statistical analysis. Mice were randomly assigned to each group, but the experimenter was not blinded to group identity.


Female NOD (NOD/LtJ) and severe combined immunodeficient NOD (NOD.scid) mice were purchased (Charles River Laboratories) and housed in specific pathogen–free conditions. Mice were considered diabetic when blood glucose measurements were ≥250 mg/dl on two successive days as determined by a Bayer BREEZE Blood Glucose Monitoring System (Bayer). All procedures were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at San Raffaele Institute, Milan (IUCAC 416 and 604). All treated mice were administered LV by intravenous tail vein injection (TU per mouse ranging from 5 × 108 to 10 × 108). Anti-CD3ε [2-C11 F(ab′)2] from BioExpress was administrated intravenously at the indicated doses.

Adoptive cotransfers

Splenocytes prepared from diabetic NOD donors (2.5 × 106 or 5 × 106) were injected intraperitoneally into 8-week-old NOD.scid mice with the indicated population at 1:1 ratio. CD4+CD25+ depletion was performed by negative selection of CD4, followed by positive selection of CD25+ cells. CD4 splenocytes were then combined with the CD4+CD25 fraction to finally reinfuse splenocytes CD4+CD25+ depleted.

Statistical analyses

Statistical analyses were performed using GraphPad Prism software. The incidence of diabetes was compared by Kaplan-Meier log-rank test. χ2 test was used to compare levels of infiltration between experimental groups. Two-way ANOVA followed with a Bonferroni posttest was used to determine statistical differences between multiple experimental groups. Findings were considered significant with values for P ≤ 0.05.


Materials and Methods

Fig. S1. InsB9–23–specific CTL response is induced after ICLV.PGK and ICLV.PGK.142T treatment.

Fig. S2. InsB15–23–specific CD8+ T cells are functional in ICLV.ET.InsB9–23.142T–treated NOD mice.

Fig. S3. Genes related to immune regulation and FoxP3+ Tregs are up-regulated after ICLV treatment.

Fig. S4. TH1, TH2, and TH17 T cell subsets are not expanded by ICLV.ET.InsB9–23.142T treatment.

Fig. S5. Depletion of FoxP3+ Tregs from total splenocytes.

Fig. S6. CD8+ T cell response to IGRP is reduced in ICLV.ET.InsB9–23.142T–treated NOD mice.

Fig. S7. Transgene-specific CTL response is induced after IDLV treatment.

Fig. S8. Frequency of FoxP3+ Tregs is increased in pancreatic islet infiltration after IDLV treatment.

Fig. S9. IDLV treatment induces expression of FoxP3+ Treg–related genes.

Fig. S10. Efficacy of ICLV.ET.InsB9–23.142T treatment is reduced in hyperglycemic NOD mice, and it is abrogated in overt disease.

Fig. S11. Definition of the suboptimal and noneffective dose of anti-CD3 mAb.

Supporting material to Fig. 3A

Data tables

Reference (71)


  1. Acknowledgments: We thank S. Gregori [San Raffaele Telethon Institute for Gene Therapy (HSR-Tiget)] and A. Valle and M. Battaglia (San Raffaele Diabetes Research Institute) for helpful scientific discussions, A. Innocenzi and A. Vino (Pathology Unit, Department of Oncology, IRCCS San Raffaele Scientific Institute) for technical assistance, and A. Nonis for statistical consulting (CUSSB, University Center for Statistics in the Biomedical Sciences). Funding: Supported by Fondazione Telethon (TIGET E1 to M.G.R. and D3 to L.N.), Italian Ministry of Health (RF-2009-1501881 to M.G.R.), and JDRF innovative grant (JDRF 47-2013-575 to M.G.R. and A.A.). Author contributions: M.A. and K.S.G. designed and performed experiments, analyzed data, and contributed to the writing of the first version of the paper. A.C. performed experiments and analyzed data. F.R. performed experiments. F.S. supervised histopathological studies. L.N. coordinated the work and revised the paper. A.A. designed experiments, analyzed data, supervised the work, and wrote the paper. M.G.R. analyzed data, supervised the work, and wrote the paper. Competing interests: The authors declare that they have no competing financial interests. A.A., A.C., L.N., and M.G.R. are inventors on a patent application owned by Fondazione Telethon and San Raffaele Hospital describing the tolerogenic IDLV.ET.142T.
View Abstract

Navigate This Article